Add libbacktrace fallback for stack traces on musl/Alpine

[hanxizh9910]

Problem

On Alpine Linux (musl-based), execinfo.h is not available, so Valkey produces no stack trace on crash. This makes debugging crashes on Alpine very difficult.

Solution

Add libbacktrace as a fallback for stack frame collection on systems without execinfo.h.

• src/config.h: Split HAVE_BACKTRACE (generic capability) from HAVE_EXECINFO (execinfo-specific). When USE_LIBBACKTRACE is defined and HAVE_EXECINFO is not, HAVE_BACKTRACE is still set so the crash handler compiles in on Alpine.
• src/debug.c:
  • Add valkey_backtrace() using backtrace_simple() as a drop-in replacement for backtrace() on musl. Macro-aliases backtrace() to valkey_backtrace() so no other code needs to change.
  • Fix fallback messages in symbolizeWithLibbacktrace to correctly distinguish between "execinfo fallback available" and "no fallback available".
• src/server.c / src/server.h: Initialize backtrace_state once at startup via initLibbacktraceFrameState() to avoid calling malloc from the signal handler, which could deadlock if a crash occurs while the allocator lock is held.

Notes

• Companion container PR: Added libbacktrace support for alpine valkey-container#139

Testing

Built on Alpine 3.23 with USE_LIBBACKTRACE=yes and triggered a crash via DEBUG SEGFAULT. Stack traces are produced correctly for all threads.

• After the implementation:

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.64%. Comparing base (678a06d) to head (71ae67a).
⚠️ Report is 1 commits behind head on unstable.

Additional details and impacted files

@@            Coverage Diff            @@
##           unstable    #3581   +/-   ##
=========================================
  Coverage     76.63%   76.64%           
=========================================
  Files           160      160           
  Lines         80472    80472           
=========================================
+ Hits          61668    61674    +6     
+ Misses        18804    18798    -6     

Files with missing lines	Coverage Δ
src/debug.c	54.83% <100.00%> (ø)
src/server.c	89.50% <ø> (-0.11%)	⬇️
src/server.h	100.00% <ø> (ø)

... and 21 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[rainsupreme]

This looks like a good change! Thanks for the fix 😊 I just had a couple ideas to simplify/document a little more.

I'm wondering why the stack trace in your testing was so short. Normally I'd expect it to be somewhat longer. Maybe if you built with CFLAGS="-g" there would be more info. I don't think this is too important though.

[rainsupreme]

It might be worth adding a comment here explaining that we preallocate bt_frame_state because allocating is unsafe while handling an async signal (such as crash handling). This probably isn't immediately obvious and might help out the next person to read this code

[rainsupreme]

When USE_LIBBACKTRACE is enabled, we should use backtrace_simple() from libbacktrace for frame collection unconditionally — not just when execinfo.h is missing. Both call _Unwind_Backtrace() under the hood, so there's no benefit to keeping the glibc path as a special case. This simplifies the guard here to just #ifdef USE_LIBBACKTRACE