Initial implementation of resource type validation

[pl-semiotics]

This adds a relatively direct OCaml implementation of the current formalisation (essentially that given in #101), focusing on the rules around how resource types are treated as creating quantified component types.

It is designed to layer cleanly around the Core WebAssembly reference interpreter, but presently requires minor changes to expose certain parsing-related information and entrypoints.

[lukewagner]

Sorry for the slow reply, but excited to check this out! Could you perhaps add a README.md to the interpreter/ directory giving step-by-step instructions (for Linux and MacOS, ideally) to go from "I've checked out this repo" to "I'm running the test suite"?

[lukewagner]

One other thing that would be super-useful is to see if this reference interpreter can run the tests in wasm-tools/tests/local/component-model and wasmtime/tests/misc-testsuite/component-model. The error strings won't match of course, so we can start by ignoring error string equality. But once the reference interpreter agrees, we'll get a bunch of coverage from importing them into the reference test suite.

[lukewagner]

Lots of great tests in there! Just a few nits and a question.

[lukewagner]

Might be nice to distinguish this case in the error string (e.g. "Cannot outer alias resource types across components" or something)

[lukewagner]

It seems like this test is missing a final part that, e.g., requires one and two to have the same function type.

[lukewagner]

Perhaps also add an invalid testcase that tries to import one and two with the same resource type?

[lukewagner]

I would think that the problem here was the lack of ascribed type, but if you did ascribe a (sub resource) type, it would be valid, and thus the error of "may not refer to non-imported uvar" doesn't make sense to me.