Tail call signature equality check only really needs to be subtyping

[compiler-errors]

We have a check for become f() tail calls which enforces that the callee's signature is compatible with the caller's signature.

Right now that compatibility check enforces equality of arguments, but I think it only should require a subtyping relationship between arguments.

I'm just opening this issue to track this so we don't forget that we should relax this and also reason about why it's sound.

[WaffleLapkin]

I think the way to justify it is as follows:

fn(A...) -> Ar <: fn(B...) -> Br implies that you can coerce fn(A...) -> Ar to fn(B...) -> Br which implies that the former can be called with the ABI of the latter (which I think really implies that their ABIs are exactly the same, but I'm not totally sure). So I think for tail calls it's sufficient to assert that callee <: caller (callee is a subtype of the caller)