feat(profiling): add Linux CPU timer stack profiler - #18724

## Description

Pure refactor of existing stack sampler internals, with no intended behavior or internal contract changes, in preparation for timer_create based cpu time profiling in #18724 

Specifically:

- Extracts the existing task/greenlet/thread-stack rendering branch in `ThreadInfo::sample()` into a local `render_unwound_stacks()` helper.
- Extracts duplicate one-time thread registration failure logging in `Sampler::register_thread()` into a local helper.

This does not introduce CPU timer profiling, new configuration, new native bindings, or changes to thread registration semantics.

## Testing

- `scripts/lint cformat`

Attempted a targeted profiling test run from the worktree, but the test runner failed during riot venv setup before executing tests due the worktree using `/home/bits/project/.riot`, which was not writable/available in that context.

## Risks

Low. This is intended to be behavior-preserving refactoring only.

## Additional Notes

No changelog needed, internal refactor only.


Co-authored-by: taegyun.kim <taegyun.kim@datadoghq.com>

## Description

The profiling safe-copy helper uses a per-thread alternate signal stack for its `SIGSEGV` and `SIGBUS` handlers. When an alternate stack was already installed by the application, CPython's faulthandler, or crashtracker, `ThreadAltStack` reused it but still disabled it during thread-local teardown. That could leave the original owner without its alternate stack and weaken its fault handling.

Track whether `ThreadAltStack` allocated the mapping itself. Teardown now disables and unmaps only profiler-owned stacks, while adopted stacks remain installed for their original owner. The installation failure path also unmaps a profiler allocation when `sigaltstack()` fails.

A release note documents the fault-handling fix.

## Testing

- Passed locally: `ddtrace/internal/datadog/profiling/build_standalone.sh -- RelWithDebInfo stack_test`
  - 4/4 native CTests passed
  - verifies that an adopted alternate stack remains installed after `ThreadAltStack` teardown
  - verifies that a profiler-owned alternate stack is disabled during teardown
- `scripts/lint cformat`

## Risks

Low. The change narrows teardown to resources owned by the profiler and does not alter public APIs or configuration. Profiler-owned alternate stacks retain their existing cleanup behavior.

## Additional Notes

Stacked out from the CPU timer profiler work, #18724, to keep this native safety fix independently reviewable.

Follow-up PR #19026 handles the complementary case where another component replaces a profiler-owned alternate stack after the profiler installs it.

[PROF-14213](https://datadoghq.atlassian.net/browse/PROF-14213)


[PROF-14213]: https://datadoghq.atlassian.net/browse/PROF-14213?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ

Co-authored-by: taegyun.kim <taegyun.kim@datadoghq.com>

## Description

Strengthens the existing uWSGI `--lazy-apps` test without `--master` by checking the PID returned from `waitpid()` and rejecting signal-based termination.

This closes a test blind spot where the process could crash during shutdown while the test continued to validate generated profile samples. The sibling worker is cleaned up before the captured exit status is asserted, including on the failure path. The change is extracted from #18724 because it is independent of the CPU timer profiler.

## Testing

- `scripts/run-tests --venv 1ef9287 -- -s -- -k test_uwsgi_threads_processes_no_primary_lazy_apps`
  - `1 passed, 12 deselected` on Python 3.13.13 with `uwsgi<2.0.30`
- `scripts/lint fmt -- tests/profiling/test_uwsgi.py`
- `scripts/lint checks`
- `git diff --check`

## Risks

None. This only makes an existing test detect process crashes that it previously overlooked.

## Additional Notes

No release note is needed because this is a test-only change.



Co-authored-by: taegyun.kim <taegyun.kim@datadoghq.com>

## Description

Strengthens the existing uWSGI `--lazy-apps` test without `--master` by checking the PID returned from `waitpid()` and rejecting signal-based termination.

This closes a test blind spot where the process could crash during shutdown while the test continued to validate generated profile samples. The sibling worker is cleaned up before the captured exit status is asserted, including on the failure path. The change is extracted from #18724 because it is independent of the CPU timer profiler.

## Testing

- `scripts/run-tests --venv 1ef9287 -- -s -- -k test_uwsgi_threads_processes_no_primary_lazy_apps`
  - `1 passed, 12 deselected` on Python 3.13.13 with `uwsgi<2.0.30`
- `scripts/lint fmt -- tests/profiling/test_uwsgi.py`
- `scripts/lint checks`
- `git diff --check`

## Risks

None. This only makes an existing test detect process crashes that it previously overlooked.

## Additional Notes

No release note is needed because this is a test-only change.



Co-authored-by: taegyun.kim <taegyun.kim@datadoghq.com>

## Description

Consolidates the two native frame updates performed for each greenlet switch into one private `record_greenlet_switch()` call and one `greenlet_info_map` mutex acquisition.

The existing behavior is preserved: the origin frame is always updated, while the running target frame is left unchanged when it is needed for parent-greenlet stack unwinding.

This is an independently mergeable prerequisite extracted from #18724.

## Testing

- `scripts/lint fmt -- ddtrace/profiling/_gevent.py tests/profiling/test_gevent.py tests/profiling/collector/test_stack.py`
- `scripts/lint cformat`
- `scripts/lint profiling-native-check`
- `scripts/lint checks`
- `git diff --check`
- Fresh CPython 3.12 native extension build
- Native linked-parent switch regression covering preserved and updated target frames
- Existing high-cardinality greenlet switch contention regression

### Benchmark

The benchmark isolates the native operation changed by this PR. Each measured run performs 1,000,000 logical greenlet switch updates after 10,000 warm-up updates:

- baseline `e11d640110`: two `update_greenlet_frame()` native calls and two mutex acquisitions per update
- candidate `93702898f8`: one `record_greenlet_switch()` native call and one mutex acquisition per update

Both `_stack` extensions were built in Release mode from the exact commits, then run in the same container pinned to CPU 0 with `--cpuset-cpus=0`.

Results over five runs:

- before: 0.531681 seconds median, runs `[0.528339, 0.531681, 0.532128, 0.534566, 0.529234]`
- after: 0.366206 seconds median, runs `[0.380415, 0.368066, 0.360803, 0.362898, 0.366206]`
- change: approximately 31% faster for the isolated native update

Environment:

- AWS KVM VM, Intel Xeon Platinum 8175M at 2.50 GHz
- 16 vCPUs, 8 cores with 2 threads per core, 61 GiB RAM
- Linux 6.8.0-1055-aws, x86-64
- Docker server 29.5.2, container pinned to one CPU
- CPython 3.12.13, GCC 14.2.0, glibc 2.41
- container image `sha256:ce7c46dbb7f07d352aecd756e19fd7f39550a72d4a13e68d30f02f5448eebd69`

<details>
<summary>Benchmark script</summary>

```python
import statistics
import time

from ddtrace.internal.datadog.profiling import stack


ITERATIONS = 1_000_000
RUNS = 5
ORIGIN_ID = 101
TARGET_ID = 102

stack.track_greenlet(ORIGIN_ID, "origin", False)
stack.track_greenlet(TARGET_ID, "target", False)

if hasattr(stack, "record_greenlet_switch"):

    def update():
        stack.record_greenlet_switch(ORIGIN_ID, False, TARGET_ID, None, True)

else:

    def update():
        stack.update_greenlet_frame(ORIGIN_ID, False)
        stack.update_greenlet_frame(TARGET_ID, None)


def run_once():
    start = time.perf_counter()
    for _ in range(ITERATIONS):
        update()
    return time.perf_counter() - start


for _ in range(10_000):
    update()
runs = [run_once() for _ in range(RUNS)]

stack.untrack_greenlet(ORIGIN_ID)
stack.untrack_greenlet(TARGET_ID)

print(f"median={statistics.median(runs):.6f}s runs={runs}")
```

</details>

This is an isolated native-call benchmark, not an end-to-end application throughput claim. The earlier 67% figure was discarded after reproducing with exact baseline and candidate builds because the candidate run had not activated the profiling hook.

## Risks

Low. The changed Python and native APIs are private and ship together. Existing parent-greenlet frame retention remains unchanged.

## Additional Notes

The GIL release around native state mutation is retained intentionally. It was introduced by #14852 in commit `54a3a0ea35` to prevent potential thread-pool deadlocks while waiting for profiler mutexes. This change reduces two such release windows per greenlet switch to one.

No release note is needed because this is an internal performance optimization with no user-facing API or behavior change.






Co-authored-by: taegyun.kim <taegyun.kim@datadoghq.com>