[win] Don't build a custom nanosleep() when not needed

[dimula73]

Instead of relying of the includes order we should properly
check if the function is actually provided by our version
of pthread library.

The patch basically reverts an original hack to resolve the
issue in msys2 and implements a proper detection mechanism
for the presence of nanosleep() in the pthread library

[dimula73]

There is still an unorthodox usage of WIN_PTHREADS_TIME_H in mlt_types.h, which dates back from 2015. Theoretically, it could also be replaced with HAVE_PTHREAD_NANOSLEEP, though we would then have to add a public-interface definition for libmlt, like that:

target_compile_definitions(mlt PUBLIC "HAVE_PTHREAD_NANOSLEEP=${HAVE_PTHREAD_NANOSLEEP}")

Though I'm not sure if such a change is acceptable or not

[ddennedy]

unorthodox usage of WIN_PTHREADS_TIME_H in mlt_types.h

It used to exist even if not in a standard. Google search on it first hit leads me to;
https://github.com/tonytheodore/winpthreads/blob/master/include/pthread_time.h

Since it is unorthodox it is OK to change it IMO.

[ddennedy]

I approve the change but wait to see what you do about mlt_types.h