Implement Memoization to Speed Up Textarea Rendering

[wesen]

This pull request introduces a significant performance improvement to the
textarea rendering process by implementing memoization to cache the results of
the wrapping computation. This change reduces the computational overhead
associated with repeated calculations, particularly for larger text areas.

This at least partially fixes charmbracelet/bubbletea#831
although insertion and rewrapping are still "suboptimal". I think it does the job
quite nicely in practice.

Key Changes:

1. Introduction of memoization.go: This new file contains the
   implementation of a generic memoization cache with a Least Recently Used
   (LRU) eviction policy. The cache is thread-safe and can be used with any type
   that implements the Hasher interface.

2. Addition of MemoCache to Model struct: The Model struct in
   textarea.go now includes a MemoCache instance for caching the results of
   the wrapping computation.

3. Modification of wrap function calls: All calls to the wrap function
   in textarea.go have been replaced with calls to the new memoizedWrap
   function. This function checks the cache for a precomputed result before
   falling back to the wrap function if necessary.

4. Addition of WrapInput struct and Hash method: The WrapInput struct
   and its associated Hash method have been added to facilitate the hashing of
   inputs to the wrapping computation for cache storage.

Important: I reused my generic Memoization which bumps go to 1.18. I'm not sure how consequent it is to introduce that kind of change incidentally, but didn't want to remove the generics for now. I can "instantiate" it all to string if sticking with 1.17 is desired.

Disclaimer: I'm pretty bullish on LLMs for programming. The memoization code, the fuzzer and the unit tests are heavily copiloted and GPT4-ed. I haven't tested the concurrency of the data structure explicitly, I'm happy to add a bunch of tests and fuzzers for that as well.

[nadimkobeissi]

Please merge this!

[wesen]

Please merge this!

You can try to use https://github.com/go-go-golems/bobatea which is my alternative version, until it's merged. I haven't tried it out outside of my go workspace, so I hope it works for you.

[maaslalani]

Hi! Thank so much for this PR, I am going to test this out very soon so we can get this in! We really appreciate the work on this one.

[meowgorithm]

Note that this would close #456.

Also note that this bumps Bubbles from Go v1.17 to v1.18 — which I don't think is an issue, but let us know if you feel differently @muesli.

[maaslalani]

Really solid work @wesen. Thank you so much for these improvements! I noticed a ton of performance benefits with this.

One thing I did change (when testing with 1000 lines of text) it was still slow until I changed the cache capacity to match the number of lines. Instead of hardcoding this, I make it so that the capacity matches the MaxHeight of the textarea so we can support the performance at large MaxHeights as well.