Implement where clauses

[brson]

Tracking rust-lang/rfcs#135.

Task list:

• PR Implement parsing and conversion of generalized where clauses #20002: Parse and process where clauses not attached to parameters.
• Where clauses that do not involve parameter types should be rejected #20019: Reject where clauses that do not mention a type parameter.
• Check that the bounds of impl are implied by trait in a more general way #20020: Check that the bounds of impl are implied by trait in a more general way.
• Parse and accept outlives expressions in where clauses #20023: Parse and accept outlives expressions in where clauses.
• Permit impl methods whose bounds cannot be satisfied to have no body #20021: Permit unsatisfiable methods to be omitted.
• Permit higher-ranked qualifiers to enclose the self type #20022: Permit higher-ranked qualifiers to enclose the self type
• Tracking issue for type equality constraints in where clauses #20041: Accept type equality constraints in where clauses

[pnkfelix]

P-backcompat-libs, but not 1.0.

[nikomatsakis]

I've been working on this somewhat in the background. The patch is somewhat more involved than I imagined initially because where clauses are more powerful than I initially imagined -- in particular, there are lots of bits of code that assume that there is an intrinsic set of bounds associated with a specific type parameter, whereas where clauses make those things scope dependent. We've been steadily removed them and by now all the deep connections are almost entirely eradicated (@pcwalton's work in #17864 is, I think, the last bit).

In my branch, I was removing all the lists of bounds from type parameters and adding in a list of predicates instead. It occurred to me today that it might also be possible to phase support for this in faster by doing it in two stages: first, leaving the code as is but adding a generalized list of predicates, and then removing the bounds lists. I'm not sure which is better.

[jroesch]

@nikomatsakis I'm looking into this after our conversation this morning. Let me know if there are any other tracking issues I should be following. I am going to go dig up your branch and work on it during the afternoon session.

[jroesch]

@nikomatsakis would it be possible to update the check list at some point? would be good to check at glance what remains. It is fine if it happens after #20020 lands.

[aturon]

Nominating for removal from P-backcompat-libs; enough has landed for us to already roll this out.