Stabilize Linux CI crash-path verification in test_stack_context

[Copilot]

The Linux Actions job failed because test_stack_context did not consistently crash under Release optimization, so test_stack_output/sentinel.txt was never produced. This broke the workflow’s sentinel-based verification step.

• Root cause

  • tests/test_stack_context.cpp used a compile-time null dereference pattern that can be optimized away in Release builds, allowing normal process exit.

• Change made

  • Reworked the crash trigger to use a runtime-loaded volatile address so the fault path is preserved under optimization.
  • Added <cstdint> and introduced a volatile global crash address used by crashSiteFunction().

• Impact

  • test_stack_context now deterministically exercises CrashCatch on Linux Release builds, restoring sentinel generation expected by CI.

#include <cstdint>

static volatile std::uintptr_t gCrashAddress = 0;

CRASHCATCH_TEST_NOINLINE void crashSiteFunction() {
    auto* ptr = reinterpret_cast<volatile int*>(gCrashAddress);
    *ptr = 42; // guaranteed fault path in optimized builds
}