ref #14972
ref #10328

Manager took ownership of logic

The callbacks return nothing (unlike drain), so they can be treated
identically regardless of the ES category.

So long as the flag is only polled, this is a little bit simpler.
(Though ideally, it would probably not be polled.)

This restores the ability to flush the decoder while it is paused.

This works around a deadlock in the ES output. When paused, buffering
cannot be measured since output latency is unknown. Even decoding
latency would be unrepresentative.

The paused flag is no longer protected by the owner lock (the FIFO lock
instead), and it is no longer useful to clear it at exit.

Edited-and-...
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Test media subitems when parsing a directory via a path, via a fd and via
various locations.

The decoder can be paused by the decoder owner while the decoder thread
is decoding. We still need to queue the last decoded picture(s) to the
(not yet paused) video output.

The "run-time" parameter was measured in terms of how long the input
thread had actually been playing, i.e. the real time since the thread
was created minus the real time spent in paused state.

This is a bit odd. It does not make much sense when transcoding, and
also seems rather counter-intuitive when playing a non-nominal rate.

This patch changes the run-time variable to match the demuxer time, as
does the start-time and the stop-time.

Playing FIFOs and devices (especially character devices) typically
only makes sense if explicitly requested. Playing FIFOs will usually
lock up. Playing devices could have any effects.

Add an option to restore the old behaviour in case someone wants it.

::facepalm