Fix subst_externtype matching on incorrect argument types

[raoxiaojia]

The substitution function for externtype is incorrectly using the substitution for deftype for function externs instead of the intended typeuse, missing the other two cases of typeuse (most importantly typeidx).

The reference interpreter is correct by using subst_typeuse.

The AL interpreter seems to be unaffected because iirc only deftypes exist during actual runtime, and subst_typeuse simply dispatches deftypes using subst_deftype, so this type error is hidden by implicit coercions (either in IL or elaboration), taking the substitution function as deliberately only defined on func externs that are deftypes. Is there something we could do in any of the checks to produce a warning for this kind of scenario (which would have made it obvious)?

For the other part of the tests, module validation should have been affected if generated from the spectec DSL, but that is out of the scope of AL as it focuses on generating the interpreter only and delegates validation etc to the reference interpreter, which is correct on this. Latex backend doesn't seem to print the full substitution definition either. As a result, the prose seems to be the only one that makes this change observable.

There also seems to be some test mismatches under spectec/test-interpreter (I think this is the one for AL?), but those are pre-existing in the current main (probably after #2093 looking at the errors).

[rossberg]

Thanks!

[rossberg]

You'll need to sync wasm-latest. (Tip: make sync-diff in the specification dir usually does the trick.)

[raoxiaojia]

My local dune test is somehow caching builds a lot for some reason -- this should fix all the test expectations.