Test case backtraces report stack from wrong threads

I am seeing a few reports in sound-open-firmware where the stack trace points to a crash in `pthread_create` but when I reproduce in lldb it points to a proper location in the code base. This is fine for reliable test cases but for unreliable ones this is problematic as the automated backtraces are the main source of information for debugging.

Example test cases
[1](https://oss-fuzz.com/testcase-detail/5536545493286912)
[2](https://oss-fuzz.com/testcase-detail/6459895332077568)
[3](https://oss-fuzz.com/testcase-detail/5904359311867904)

Start by reproducing the automated backtrace behavior using the three linked OSS-Fuzz test cases and compare it with the lldb results described in the report. Trace how crashes involving pthread_create are collected and identify why the reported stack points to the wrong thread; done means the automated backtrace identifies the relevant crash location for unreliable test cases.

Written by the indexing model from the issue text.