Episode 13b: Performance improvements

[obiwac]

This episode is split into two parts, and most of this was prototyped in my mcpy-accelerated repo:

• EP13a: General cleanup (done by Formatting & linting #107 and Episode 13a: General cleanup #108).
• EP13b: Performance improvements (in terms of loading time).
• EP13c: Performance improvements (in terms of frame time).

This PR focuses on the performance part. This is what needs to be done:

• Attach profiler (in practice, comment it out in this repo no need and add it as a dev dependency in pyproject.toml cProfile comes with Python).
• Set up buildsystem with Poetry's unstable build.py (but looks stable enough for me).
• Speed up chunk generation (cythonize, C code for the really sensitive bits).
• Do save chunk caching (as nbtlib is now dominantly slow vs. chunk mesh generation).
• Load chunks progressively (now that they can be generated in less than a frame's time).
• Improve performance of entities (colliders (?) and physics engine). In practice that just means the player at the moment, even though mcpy-accelerated also does this for non-player entities (mobs). EP13c (TODO link to PR).
• CI workflow to make sure all our Cython modules are building correctly.
• Investigate if this can be integrated and closed: Chunk Mesh Generation Optimization #96. Won't make a difference.
• DON'T FORGET TO UPDATE CODE TO RE-LOAD FROM NBT IF THERE ARE CHUNK FILE CHANGES!
• GLM. EP13c (TODO link to PR).

All of this still needs to be done for community, but I'll probably do that last minute before releasing EP13b.