fix: allocation checks, a log buffer overflow, result leaks, and a per-line rebuild

[somethingwithproof]

Four small fixes in spine.c and util.c, all found while auditing classes that #542 and #557 fixed elsewhere in the tree but did not finish.

Based on current develop, after #569 and #570.

#564 — php_processes, debug_devices, db_pool_local and db_pool_remote were dereferenced on the line after calloc() without testing the result. #542 gave the two in poller.c a die(); these were missed. Same treatment, same message shape. snmp.c:1028 and util.c:124 already checked theirs.

#565 — spine_log() bounds both strncat() calls so the message may fill flogmessage exactly, then appended the newline unconditionally, putting the terminator one byte past a LOGSIZE stack buffer. Same class as issue#561. The append is now bounded.

#566 — four settings helpers returned without db_free_result() when the query reported rows but the fetch yielded NULL. Every other exit in each function frees it.

#567 — get_date_format() did a malloc/free pair, two range checks and two switches on every log line to rebuild a string that only depends on two settings read once at config load. It is now built in read_config_options(), while spine is still single threaded, and get_date_format() returns that storage. spine_log() no longer frees it. No behaviour change; the format for every code is unchanged, which the existing tests in test_safety_fixes.c already assert.

Verification. Clean ubuntu:24.04 container: builds at 4 warnings, matching develop, and make check passes all four suites. The new test is wired into the check_PROGRAMS list #570 added.

tests/unit/test_log_newline_bound.c is new and pins #565 down. It fails against the old strcat(): built with ASan and the pre-fix predicate, test_full_buffer_is_left_alone reports a stack-buffer-overflow; with the fix all five pass.

Two things deliberately not here. All fourteen mysql_fetch_row() call sites are already guarded, by the loop condition or an explicit test, so there was nothing to fix beyond the leak in #566. And the eight strcat() calls at util.c:885-914 append compile-time literals into BUFSIZE buffers that were just zeroed, well under 100 bytes total; they are safe by construction and left alone.

[TheWitness]

Merge conflicts.

[somethingwithproof]

Conflicts are cleared. GitHub reports this MERGEABLE against current develop, and #576 is closed as a duplicate of this one, so the overlap on spine.c and util.c is gone.

[somethingwithproof]

Consolidated into #597, which carries this branch's commits unchanged.

Every pair of these ten conflicted on tests/unit/test_linked.c because each appended to the same registration array, so whichever merged first would have forced a rebase on the other nine. #597 is one review and one approval for the same 20 commits, still one logical change each.

Nothing here is dropped. Reopen this if you would rather review it separately.