The previous design had a race condition. When WaveViewThreads::stop_threads() was called, it would
first acquire the mutex, then set _quit, then call condition.broadcast(). But worker threads would
check _quit without holding the mutex. It was therefore for a thread to be delayed in its
own lock acquisition by the ::stop_threads() caller, then end up back in cond.wait() AFTER
the cond.broadcast() was done. Such a thread would sleep forever and never wake up.
This new design removes WaveViewDrawRequestQueue, which was a clean encapsulation of the
queueing aspects of WaveViewThreads, but unfortunately made correct mutex acquisition
and condition signalling/waiting needlessly complex. THe mutex, condition variable
and actual queue were moved into WaveViewThreads, and all worker threads execute a method
of the class which gives the appropriate code easy access to the mutex and condition var,
which must always be used together.
"While 'atomic' has a volatile qualifier, this is a historical
artifact and the pointer passed to it should not be volatile."
Furthermore "It is very important that all accesses to a
particular integer or pointer be performed using only this API"
(from https://developer.gnome.org/glib/2.68/glib-Atomic-Operations.html)
Hence initialization of atomic variables is changed to also use
this API, instead of directly initializing the value.
This also fixes a few cases where atomic variables were
accessed directly.
see also libs/pbd/pbd/g_atomic_compat.h
