Refactor typeck code handling `self` and type parameters

[catamorphism]

I'm about to land my patch for #3979, but @nikomatsakis suggested refactoring the code related to it more thoroughly in the future, in order to:

• stop treating self specially, but rather, treat it like any other type parameter
• don't concatenate bounds from different type parameters; rather, pass around vectors of vectors of bounds

[catamorphism]

I'm working on it. Moved up to 0.6 -- it doesn't look too hard (famous last words...)

[catamorphism]

I need to talk to @nikomatsakis about this.

[nikomatsakis]

Not critical for 0.6; removing milestone

[catamorphism]

Nominating for milestone 5, production-ready

[graydon]

just a bug, removing milestone/nomination.

[huonw]

Visiting for triage; sounds like something that @msullivan might know something about.

[catamorphism]

Sadly, I never got around to this. @nikomatsakis , have you perhaps already done this as part of another patch?

[flaper87]

triage bump. There's still work to be done here. I think @eddyb did something already.

[japaric]

@nick29581 Can this issue be superseded by the planned work on UFCS (re: the self treatment part)?

[nrc]

Even with the UFCS work, self is still treated a bit specially in the compiler. It shouldn't be. So I think this issue is still relevant (maybe even more so) post-UFCS.