• Add svt-av1 option --keyint FRAME-OR-DURATION argument supporting frame integer or duration string.
  E.g. --keyint=300 or --keyint=10s.
  Default keyint to 10s when input duration is over 3m.
• Add svt-av1 option --scd true|false argument enabling scene change detection.
  Default scd on when using default keyint & input duration is over 3m.
• Add --svt ARG for additional args, e.g. --svt mbr=2000 --svt film-grain=30.
• Add --vfilter ARG argument to apply a ffmpeg video filter (crop, scale etc) to the input before av1 encoding.
  E.g. --vfilter "scale=1280:-1,fps=24".
• Add --pix-format ARG argument supporting yuv420p10le (default) & yuv420p.
• Add vmaf configuration --vmaf ARG, e.g. --vmaf n_threads=8 --vmaf n_subsample=4.
• Rename vmaf command argument --reference (was --original).
• Add vmaf command --reference-vfilter argument, similar to --vfilter.
• Default vmaf n_threads to the number of logical CPUs.
• Add --temp-dir argument to specify storage of sample data.
  May also be set with env var AB_AV1_TEMP_DIR.
• Add --sample-every DURATION argument, default "12m".
• Remove 3 sample default, this is now calculated using --sample-every 12m default.
• Create samples concurrently while encoding to reduce io lags waiting to encode.
• crf-search re-use samples for crf analysis.
• Linux: vmaf use fifo to convert both reference & distorted to yuv which fixes vmaf accuracy in some cases.
• Support multiple audio & subtitle streams.
• Use 128k bitrate as a default for libopus audio.
• Remove --aq.
• Fail fast if ffmpeg cut samples are empty (< 1K).
• Handle input durations lower than the sample duration by using the whole input as a single sample.