jemalloc: use GCC intrinsics for unwinding in profiling

[teskje]

This is a workaround for deadlocks caused by nested unwinding. Alternative to #32139.

Motivation

• This PR fixes a recognized bug.

Fixes https://github.com/MaterializeInc/database-issues/issues/9159

Tips for reviewer

Checklist

• This PR has adequate test coverage / QA involvement has been duly considered. (trigger-ci for additional test/nightly runs)
• This PR has an associated up-to-date design doc, is a design doc (template), or is sufficiently small to not require a design.
• If this PR evolves an existing $T ⇔ Proto$T mapping (possibly in a backwards-incompatible way), then it is tagged with a T-proto label.
• If this PR will require changes to cloud orchestration or tests, there is a companion cloud PR to account for those changes that is tagged with the release-blocker label (example).
• If this PR includes major user-facing behavior changes, I have pinged the relevant PM to schedule a changelog post.

[def-]

We can do a bunch of anyhow! calls to have a quick stress test to see if this prevents the issues. (or revert #32138)

[def-]

Please ignore the upgrade test failures in Nightly, they are fixed by #32265

The only thing I'm afraid of here is that we might miss the argument in some C library and then get a crash in production as a consequence of that

[teskje]

The only thing I'm afraid of here is that we might miss the argument in some C library and then get a crash in production as a consequence of that

Yeah, it's a real risk. Bazel helps us out here a lot by letting us set -fno-omit-frame-pointer as a cflag for all external libraries we build. So the risk is limited to libraries we don't build as part of the Bazel build process. At the moment, the ones I'm aware of are:

• The stdlib, which is shipped with frame pointers enabled: Enable frame pointers for the standard library rust-lang/rust#122646
• libc++, which we build ourselves, so we can also enable frame pointers there.
• Any dynamic libraries we link against (libc, libgcc_s, ...). Those are provided by Ubuntu packages, so they also have frame pointers enabled.

Some risk still remains, but I'd hope that CI will flag anything I missed above. It did point out libc++ as a library I missed initially.