Optimize sounds by removing unused ones and re-encoding with VBR

[dsnopek]

This PR aims to help somewhat with #126 by doing two things:

1. Removing unused sound files. This time I did the removal less manually, and used some scripts to try and find the used files. Hopefully, I got this right! It seems fine running the game after the removals
2. Re-encodes all OGG files with VBR. I didn't look at every file, but quite a few were fixed bit rate, rather than variable bit rate, so I did a mass conversion via ffmpeg -c:a libvorbis -qscale:a 3. Converting to VBR is theoretically a very conservative change because VBR vs fixed shouldn't have a noticeable change in quality. However, since the source file was also an OGG and it's a lossy format, just re-encoding at all will lead to some quality loss. If the original source files were available, re-encoding from them would be better

I split these two changes into two separate commits, because I'm not entirely sure about nr 2. Unless I messed something up, I think nr 1 is a pure win, however, nr 2 may represent a small loss in quality as described above

OR, we may want to consider a more extreme version of nr 2 where we reduce the sample rate (would have a noticeable difference in quality, but may not matter given the ambient nature of the audio?) or use a lower quality setting for VBR (so having a lower overall bit rate, again, with a noticeable loss in quality).

Here are the size numbers for the assets/sound directory:

• In main: 46 MB
• After nr 1: 38 MB
• After nr 2: 27 MB

So, about a ~20 MB reduction in the source files. When exporting to web, I see about a ~18 MB reduction in the size of the PCK file