grapheme popularity as hat metric

[josharian]

Checklist

• [-] I have added tests
• [-] I have updated the docs and cheatsheet
• [-] I have not broken the cheatsheet

[josharian]

@pokey @AndreasArvidsson et al, now that we record command history, we can run experiments!

this is a tweak to the hat allocator that I have been using in my local install for ages (so I'm pretty sure it's not a disaster). there are some performance improvements we would want to make before actually submitting the code, if the functionality is as good or better than the current hat allocator. (I think the code is both simpler and faster to execute than the current allocator, so the bar here is just to not be worse in real world use.)

want to install it locally and gather some stats? (it is okay if the answer is no. :P)

[josharian]

Test failures are TODOs intentionally left in the code for now.

[pokey]

I'd prefer not to break my update chain to experiment with this one. How hard would it be to feature flag it and keep track of flag value in log?