fix 234, fix 236, bugs related to parametric types

[EliasC]

The problem causing 234 was that methods from included traits did not
have their type parameters instantiated when being looked up.

The problem causing 236 was a weird call to replaceTypeVars when
typechecking new.

[albertnetymk]

One minor comment: matchMethod was introduced to abstract the logic of finding matched method. Other that that, this PR looks OK to me.

[EliasC]

I thought I'd reduce the ambiguity of the already overused concept of "matching". (==m) . mname is both readable and understandable in my opinion.

[albertnetymk]

Explicit == doesn't seem the same level of abstraction with the rest body of the function. Unifying the method matching logic is handy once method overloading is added. Both points are rather subjective.

[EliasC]

Subjectiveness is ruining all the fun... Reverted now.

[albertnetymk]

Sorry. My point is that it's subjective, so the author of PRs should hold the right to make the judgement call.

I am OK with the PR now. Merging?

[EliasC]

I actually agree with you now (otherwise I wouldn't have changed it back). The overloading argument makes sense!

[albertnetymk]

Merging it in 5 mins if no one objects.

[kikofernandez]

looks good to me