Error and retry chunk if frame count mismatches

[shssoichiro]

I can't think of a situation where someone would want this issue to be
ignored, so we should treat it as any other encoder error and perform 3
retries, then exit if it continues to fail. This probably indicates some
kind of issue with decoding or applying filters to that chunk.

[redzic]

I think we also have to explicitly decrement the progress bar in the case of that type of failure.

[redzic]

Perhaps this should be a warning message instead of an error since we are still restarting the chunk? I think this might be confusing to some users, what do you think though?

[shssoichiro]

That's fair, it looks like in the other places we warn unless it's at max retries and bails.

[shssoichiro]

I happened to have a file that got through the whole encode with one chunk with a significant frame mismatch. It muxed and happily deleted all of my temp files, and the resulting output had a significant desync after the broken chunk. So, just confirming that retrying after this and erroring after 3 failures is the correct approach.

[redzic]

Looks much better without the code duplication, thanks! My only suggestion is to change line 250 of broker.rs to say "encoder failed" instead of "encoder crashed", to both match our warning message and to reflect our now more generic usage of EncoderCrash.