feat: Lazy wasm pt4

[Thunkar]

Hopefully one of the final parts of this refactor. Handles all the hashing done through bb.js. This is a monster and I'm sorry, but if there's something we do it's hash stuff.

There should not be any unhandled promises left by this PR, thanks to eslint and careful reviewing anywhere we could now have somethingThatIsNowAPromise.toString().

UX regressions introduced by this PR:

• Authwitness interactions in public are now async, so they cannot be chained. Will try to improve in the future, can be seen in e2e_authwit.test.ts
• Function selectors now have to be computed asynchronously, which is awkward since we used to have getters for them. I think this can be easily done in codegen and will try to improve in the future.
• Same with TxHashes, but those are more complicated to precompute

Other considerations:

• I had to use parseAsync with zod instead of parse, which is known to be slower. I don't think this is going to be a big issue, since it's only used when doing RPC calls, but would love some inputs CC/ @spalladino
• @trackSpan decorator can now take an async function for its attributes
• Performance should not suffer since we were crossing the WASM boundary anyways even when using the sync version. In some cases, the usage of Promise.all could have potentially improved it.
• I have kept the sync version of hashes under @aztec/foundation/crypto/sync, so that @aztec/merkle-trees wouldn't complain. Since the TS implementation of our merkle trees is likely going away, I didn't bother updating them.

[dbanks12]

AVM related stuff all looks good!