We want to update format and decoder without having to
es_out_Add/es_out_Del which would alter id and selected state.

logs are usually spammed on slow buffering.

This reverts commit 6aeaeaa02239a0cb6474c2504c1d525b006c41bd.

not related to the other pace controls

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

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

Fix the workaround in AudioTrack_ResetPlaybackHeadPosition. Don't wait for an
unknown time that the position stabilizes but reset the AudioTrack object.

The audiotrack head position shouldn't be ahead of the module position anymore.

Refactor AudioTrack creation, and add a function that destroy and recreate an
AudioTrack object using the previous arguments.

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

This is not needed. And for the normal mac builds, we even do not
have this library at all.

close #13838

There are still many cases where waking up the thread sleeping is only
half of the problem and waking it up half of the solution. For instance,
the net_* and stream_* I/O functions have no ways to return a non-fatal
error.

This reduces the system call overhead.