Add checker conformance tests for type parameter deduction for type(t) - #529

could you add explicit tests for the edge cases around free parameters?

I don't think these are super practical expressions but maybe better to document the weirdness than not?

It's also a bit of a can of worms so it might be fine to leave unspecified for now until we sort out the other issues with type unification.

I think the updates in C++ and java have us with something like:

[type([]), int]  -> list(dyn)

[] + [[type(1)]] + [[type([])]]  -> list<list<dyn>>

false ? int : type([]) -> no overload

Let me send these out as a separate PR, the [] + [[type(1)]] + [[type([])]] case especially requires further discussion (all of the stacks are divergent, so we need to decide on a behavior)

I seem to recall empty iteration ranges within nested comprehensions have an odd effect on type unification as well