Fix parallel access crashes and misbehavior

[jmid]

Parallel usage is memory unsafe (read: may crash) as documented in #120, ocaml/ocaml#11607, and ocaml/ocaml#13046.

This PR goes for the simplest possible fix: adding a single global lock by dusting off the first commit of https://github.com/dra27/flexdll/tree/sledgehammer and suitable rebasing, renaming, and error handling.
Author credit thus goes to @dra27 - any errors are mine.

For the error handling, I've tried to make it fit with @shym's TLS-based error handling from #112.
I'm unsure how to test these error code paths though without explicitly mocking with the source code to create an invalid lock handle.

With the fix

• the ocaml/ocaml testsuite passes incl. the disabled tests/lib-dynlink-domains test from Temporarily disable the lib-dynlink-domains test on Windows ocaml#11607
• the reproducer from Parallel Dynlink usage under Cygwin+MinGW is unsafe ocaml#13046 also passes
• the Dynlink stress test from multicoretests passes

(these have been tested under MinGW in a Cygwin-shell)

[MisterDA]

Suggested change

	if (WaitForSingleObject(units_mutex, INFINITE) ==WAIT_FAILED) {
	if (WaitForSingleObject(units_mutex, INFINITE) !=WAIT_OBJECT_0) {

This would cover the improbable (impossible?) case of the mutex having been abandoned.
https://learn.microsoft.com/en-us/windows/win32/sync/using-mutex-objects

[jmid]

I ended up reverting this one.

WAIT_ABANDONED does not offer GetLastError-integration, so I started adding an explicit error message case for it - only to then hit it during testing.

I therefore suppose these two cases were written to handle both WAIT_OBJECT_0 and WAIT_ABANDONED as success-cases (WAIT_TIMEOUT should not happen with INFINITE...)

[dra27]

My reasoning I think is that WAIT_ABANDONED should be treated like success?

[dra27]

(i.e. I agree with @jmid's assessment!)

[MisterDA]

Suggested change

	if (WaitForSingleObject(units_mutex, INFINITE) ==WAIT_FAILED) {
	if (WaitForSingleObject(units_mutex, INFINITE) !=WAIT_OBJECT_0) {

[jmid]

(same for this one)

[MisterDA]

I've tested an finally understood why the code is correct, thanks Jan! I think this is good to go.
Minor suggestion still.

[jmid]

I've addressed the last minor comment, added a CHANGES entry, and rebased on master.

[dra27]

Thank you both for your work on this, and sorry for taking quite so long to merge it!