HPC-GAP: crashes related to thread control

[fingolfin]

While working on PR #3741, I discovered various ways to crash or lockup HPC-GAP. According to @rbehrends these seem to be due to thread control not being prepared for a thread being paused/killed/interrupted so quickly. While the concrete examples below are of course artificial, I believe similar things can happen in realistic code, and then would be difficult to track down.

Here are the problematics snippets; a crash or deadlock happens in WaitThread. And even if GAP manages to get past them, something in the internal state seems to get broken, leading to weird behaviour and lockups later on.

My tests were done on macOS 10.14, but Travis also had issues, so I don't think this is OS specific.

t := CreateThread(function() MicroSleep(100); end);;
PauseThread(t);
ResumeThread(t);
WaitThread(t); # FIXME: sometimes deadlocks

t := CreateThread(function() MicroSleep(100); end);;
InterruptThread(t, 17);
WaitThread(t); # FIXME: crashes in pthread_join

t := CreateThread(function() MicroSleep(100); end);;
KillThread(t);
WaitThread(t); # FIXME: crashes in pthread_join