Effeciently extract subset of images from video

[SeeRich]

Related to this issue: #83

FFmpeg CLI has the ability to perform frame accurate input seeking which is very fast if you want to extract a frame at a time. See: https://trac.ffmpeg.org/wiki/Seeking.

However, I have a very long video (~2-3 hours) and hundreds of timestamps at which I want to extract frames. Is there a way I can leverage the input seeking at the libav level? I don't want to just call the CLI in a loop for each timestamp.