Posix StackDumpSignalHandler should re-raise signal

In-process stack dumping has caused problems on Posix platforms
due to the signal handler's behavior. Historically on some platforms
it has just called _exit(1) which is bad for tests that need to verify
a process crashed with the correct signal (e.g. sandbox tests,
crash reporter tests, assertion tests, MTE tests). More recently it
has been returning from the signal handler and waiting for the crash
to occur again, which may not re-raise the signal if the signal was
delivered asynchronously (e.g. via kill(), or async MTE faults), or
possibly due to race condition crashes.

StackDumpSignalHandler() should instead just purposely re-raise
the signal, very similarly to crashpad.

On Linux, ChromeOS, and Android we try to use rt_tgsigqueueinfo to
re-raise the signal and preserve 100% of the crashing context. On
other Posix platforms (Mac) and very old, unsupported Linux kernels,
we just call raise() again which will preserve most of the important
context of the crashing instruction.

This effectively reverts
b54981c4
e5c3e3c6

Bug: 551681
Change-Id: Ice0e3f7dd6c58ef8bc15a404c5251a15fbc0c2a0
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/4602549


Reviewed-by: Daniel Cheng <dcheng@chromium.org>
Commit-Queue: Matthew Denton <mpdenton@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1161968}