fix(unwinder): Guard invalid ARM32 instruction pointers

[jpnurmi]

On ARM32, unw_step first asks libunwind whether the current frame is a signal frame. The vendored implementation answers that by reading an instruction directly from the cursor's initial IP. Its local address-space validation is enabled only after this probe, so an unmapped IP faults inside the crash handler instead of returning an unwind error.

This surfaced when unrelated stack-layout changes in #1974 made the handler fallback produce 0x10c as its initial IP. No attachment code ran on that path; the changed layout only exposed the unwinder's unchecked read. The resulting SIGSEGV recursively entered the crash handler and prevented the original abort from being captured.

Check /proc/self/maps before calling unw_step. Keep the initial frame, as the existing invalid-stack-pointer path does, but stop the walk before libunwind can dereference an unmapped address. Cover the failure with an ARM32 context whose initial IP is intentionally unmapped.

[codecov]

Codecov Report

❌ Patch coverage is 50.00000% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 74.35%. Comparing base (efe5fff) to head (67c492c).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1977      +/-   ##
==========================================
+ Coverage   74.19%   74.35%   +0.16%     
==========================================
  Files         104      104              
  Lines       25634    25638       +4     
  Branches     4626     4627       +1     
==========================================
+ Hits        19018    19062      +44     
+ Misses       5317     5272      -45     
- Partials     1299     1304       +5     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.