Named mutex: Use flock instead of pthread process-shared mutex in some more cases

Named mutex: Use flock instead of pthread process-shared mutex in some more cases

Workaround for dotnet/coreclr#5456:
- Sometimes, a timed wait operation is not getting released, causing a hang
- Due to the hang, it is not possible to detect this issue with code
- Temporarily disabled the use of pthread process-shared mutexes on ARM/ARM64. File locks will be used instead.

Workaround for dotnet/coreclr#5872:
- On Alpine Linux, a pthread process-shared robust mutex is detecting the case where a process abandons the mutex when it exits while holding the lock, but is putting the mutex into an unrecoverable state (ENOTRECOVERABLE) instead of assigning lock ownership to the next thread that is released from a wait for a lock and notifying of abandonment (EOWNERDEAD).
- Added a test case to detect this issue, to have it use file locks instead

Close dotnet/coreclr#5456

Commit migrated from https://github.com/dotnet/coreclr/commit/f63816767aee67509ac3c61e94ff95d371471493