ffplay: signal the frame queue before closing audio

Fixed regression caused by 631ac655c00e978e19d05dab572bc1ffd6078c63 when ffplay
does not quit if the audio thread is blocked.

Set -scan_all_pmts 1 in ffmpeg, ffplay and ffprobe if not set by user.

Fixes ticket #3762.

ffplay: fix indentation after last commit

ffplay: implement separete audio decoder thread

for more detail
http://www.spinics.net/lists/arm-kernel/msg368418.html

ffplay: remove manual bug option handling code

It is supported as a codec option, and those are already parsed.

ffplay: remove manual bug option handling code

It is supported as a codec option, and those are already parsed.

Merge commit 'e58a140cf91d1a9cdfa3115d73c923dab0e9e7dc'

* commit 'e58a140cf91d1a9cdfa3115d73c923dab0e9e7dc':
  avplay: Always free opts

Conflicts:
ffplay.c

Merge commit 'cf83c0173917a44b89a25ea2b435429bbf9ee767'

* commit 'cf83c0173917a44b89a25ea2b435429bbf9ee767':
  avplay: Always free find_stream_info options

Conflicts:
ffplay.c

ffplay: dont leave swresampler in half initialized state

On init failure, let's just free it, so next time it will be recreated from
start.

Also fixes Coverity CID 1241515.

ffplay: Separate AVFrame from AVSubtitle

This is more robust and should avoid false positives in coverity
That is if the type is ever mixed up, after this patch that would
just be a null pointer dereference instead of a random possibly
out of array access

ffplay: use av_codec_get_pkt_timebase()