Async gpu std example (#315)

## Summary

Adds `examples/async_gpu_std/`, a counterpart to
`examples/async_gpu_launch` that demonstrates the same double-buffered
async CPU+GPU optical photon pattern but **using only C++ stdlib
primitives** _ no `G4TaskGroup`, no `G4Mutex`. The GPU worker is a plain
`std::thread` driven by `std::mutex` + `std::condition_variable` + a
bounded `std::queue<GPUTask>`.

 ## What's in it

  | file | purpose |
  |---|---|
| `async_gpu_std.h` | `GPUTaskManager` (worker thread, bounded queue,
backpressure), Cerenkov + multi-component scintillation genstep paths,
`PhotonSD`,
  sync/async `RunAction` |
| `async_gpu_std.cpp` | `main()` with manual argv parsing, physics list,
action initialization |
| `CMakeLists.txt` | standalone `find_package(eic-opticks)` build (also
pulls `glm` first to work around the missing `find_dependency(glm)` in
the installed
   eic-opticks Config.cmake) |
  | `apex.gdml` | bundled detector geometry (no external references) |
| `run.mac` | `setStackPhotons false` on Cerenkov + Scintillation, 100
beam-on events |
| `run.sh` | reads inputs from the example dir; `--sync` / `--async`
(default) flags |

  ## Architecture vs `async_gpu_launch`

  | concern | async_gpu_launch | async_gpu_std |
  |---|---|---|
  | worker thread | `G4TaskGroup<void,void>` | `std::thread` |
  | mutex | `G4Mutex` / `G4AutoLock` | `std::mutex` |
  | wakeups | implicit via task group | `std::condition_variable` |
  | queue | implicit | `std::queue<GPUTask>` (bounded) |
  | backpressure | none | `GPU_MAX_QUEUE_SIZE` (default 3) |

Simulation behavior is otherwise identical _ Cerenkov + 3-component
scintillation gensteps, sphoton `.npy` hit dumps per batch.

Author: ggalgoczi

examples/async_gpu_std/CMakeLists.txt

@@ -0,0 +1,26 @@
+cmake_minimum_required(VERSION 3.20)
+
+project(async_gpu_std LANGUAGES CXX)
+
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
+set(CMAKE_CXX_EXTENSIONS OFF)
+
+# Order matters: pull in glm before eic-opticks because the installed
+# eic-opticks::SysRap target's link interface references glm::glm but the
+# package's Config.cmake does not declare it as a dependency.
+find_package(glm REQUIRED)
+find_package(eic-opticks REQUIRED)
+find_package(Geant4 REQUIRED ui_all vis_all)
+find_package(Threads REQUIRED)
+
+add_executable(async_gpu_std async_gpu_std.cpp async_gpu_std.h)
+target_link_libraries(async_gpu_std
+    eic-opticks::G4CX
+    eic-opticks::SysRap
+    eic-opticks::U4
+    ${Geant4_LIBRARIES}
+    Threads::Threads
+)
+
+install(TARGETS async_gpu_std)