Mega rewrite

[pufferfish101007]

This is a very extensive rewrite of much of the codebase. Available to test at https://pufferfish101007.github.io/hyperquark/rewrite-2024/.

Adding new blocks is now quite straightforward - prs to pufferfish101007:rewrite-2024 are welcome, especially for simpler blocks - items marked as 'easy' on the below TODO list would be good starting points

Notable changes:

• modularisation
• proper unit tests to ensure wasm (and imported js) is valid (resolves unit tests are not proper unit tests #18)
• simplifies adding new blocks
• boxed values use NaN boxing rather than tags
• ints are i32 rather than i64
• casting should always pick the best type to cast to
• clones a bit less (resolves clone is potentially overused #20)
• add compiler settings (currently mostly unused)
• auto track implemented blocks (resolves Auto-track implemented blocks #15)
• converts nans to zeros (resolves NaNs should be treated as 0 in mathematical operations #17)
• makes output types more specific (e.g. +ve/-ve, nan-ness) - based off Update improved type analysis to develop TurboWarp/scratch-vm#216
• adds own internal wasm! macro to generate code (with special pseudo-instructions for things like zeroing nans)
• github actions deploys each branch to its own gh pages subdirectory
• warped custom blocks with arguments supported
• turbo mode works properly now
• modules optimised using wasm-opt

Still TODO:

• don't grow table for each new string (https://discord.com/channels/1095993821076131950/1095993821076131953/1362702588008726560)
• control_while and control_foreach should be implemented #13? (easy)
• other operator blocks (easy) (random, mod, [] of (), round)
• pen (easy; machinery already in js)
• costume rendering/looks blocks (easy; machinery already in js)
• const folding
• setup gh pages to have 'production' build at root dir, with other builds (including at /hyperquark/) to be 'development'
• add option to disable ints
• improve variable type specialisation
• non-warped custom blocks?

Some tests are currently failing and are ignored in CI; these should be resolved with kaleidawave/ezno#121 and don't indicate problems in hyperquark itself.