Caching and other optimizations for method call resolution

[nikomatsakis]

Method call resolution takes a long time. It should not.

[nikomatsakis]

This PR lays groundwork to support caching: #18694

@pcwalton also had a clever idea for quick-reject, which could be rebased on top of that PR: #17995

[steveklabnik]

Triage: unsure if this has been implemented or not.

[nikomatsakis]

No.

[ishitatsuyuki]

It has been a long time since this issue have been around. @nikomatsakis Is this still a thing we should do?

[nikomatsakis]

Hmm. I think we can go ahead and close. We've got quick reject and we're using trait evaluation, which has a cache. There is probably still more caching that could be done but it's not where I would start.