0.5
  - Ported read_audio_bin() to pure C, match parameters with read_audio_fft()
  - Fix total_time parameter in av_audio_convert() to account for start time
  - Unit test now uses CPU time instead of elapsed time (BDR)
  - Added start_time and end_time parameters to read_audio_fft()
  - Rename av_video_info to av_media_info
  - Minor tweaks to configure script

0.4
  - Windows/MacOS: CRAN binary packages now ship with ffmpeg 4.2.2 + drivers: vorbis,
    vpx, ogg, lame, xvid, x264
  - Removed forced keyframes (gop_size) when encoding video from graphics or PNG images.
    This reduces output video size, hopefully no major quality regressions.
  - New functions for audio! read_audio_fft() and read_audio_bin()
  - New method plot.fft_data() for spectogram, e.g: plot(read_audio_fft('sample.mp3'))
  - New function av_spectrogram_video() to create spectrogram movie
  - Detect and error in configure if FFmpeg is too old
  - New parameters channel and sample_rate in av_audio_convert()

0.3
  - Input can now contain video, i.e. multiple images per file. Previously
    we would only read a single image from each input file, but now we read
    all of them.
  - Major refactoring to suppor the above, hopefully no breakage.
  - New functions av_video_convert() and av_audio_convert()
  - New function av_video_images()
  - Switch to R_UnwindProtect to implement cleanup on exit
  - Test, tests, tests, so many tests.

0.2
  - Initial CRAN release