Add typechecking compiler pass

[ecioppettini]

Changes in this PR:

Parser

Add syntax for (optional) type annotations on variables.

let x : ty = ...

Add effect annotation syntax on coordination scripts

abi Abi {
  effect E1();
  effect E2();
}

script {
  fn foo(): u32 / { Abi } {
  }
}

I feel like using { } as delimiters here makes it a bit weird because it's too similar to the function body. But I also think it's the more proper syntax since that's the usual math notation for sets.

I don't know if it's worth changing the function syntax for it though. It's also the syntax that was on the slides I think.

Add syntax to annotate yield and resume in utxo

utxo Utxo {
  Yield u32
  Resume string
}

Added just to be able to typecheck the yield inside the main body, not necessarily because it's the best syntax. Also I couldn't find examples of these on the slides. We could put the annotation on the return type of the "function", but I don't see a good way of annotating the Resume that way.

Other

1. There was a bug where loop { } was parsed as a record instead of a loop in the parser.
2. Also changed raise to not take an expression, since effects are not really expressions.
3. Added spans to expressions for errors. Still need to add some spans for other things though, some errors have a dummy span right now.

Typechecking

This adds a typechecking pass to the compiler after name resolution, with type inference for locals. We don't necessarily need to keep this restriction, since the algorithm should work anyway, but I'm not sure it's worth it.

There are still many holes though, but I tried to cover at least most of the things needed for coordination scripts. The examples that we have don't typecheck yet though. In particular with tokens I think there are some type annotations missing in the AST that need to be added before.

Open question

I'm not sure what's the proper syntax to resume a utxo with a type. Should it be utxo.resume(val)? We have a resume keyword in the syntax, but it was meant more for the effect handlers.

[SpaceManiac]

Should it be utxo.resume(val)?

This seems natural to me.

[SpaceManiac]

Minor things from first pass review, looks good overall

[SpaceManiac]

Long term, if we're looking to track the Wasm component model, we'll want I/U 8/16 and char. But this doesn't need to happen now.

[ecioppettini]

I agree with having those types, but I'm not sure we necessarily need 1 to 1 parity to the wasm component model.

[SpaceManiac]

Maybe TokenIntermediate for clarity / to avoid confusion with codegen Intermediate?

[ecioppettini]

I'll make that change. It's a bit confusing since the one in the ast it's also Intermediate. But eventually my idea was to unify both types, so we can fix it that way eventually.

To be honest I feel like Intermediate in general is not the best name to the type (in the language), I would maybe call it FreeToken , UnboundToken or something like that.