Typechecker cruft tracking issue

[SebastianMestre]

As discussed on #324, there are a lot of legacy issues in the typechecker that make it buggy and more complex than it needs to be. Here are some ideas that popped up after reading through the most important functions

• remove typefunc unification variables (Remove typefunc vars #324)
• expose a better typevar constraint API, replacing the new_dummy_* family (Type constraint api #329)
• check typevar constraints when unifying a concrete type -- right now satisfies() just returns true
• expose better, shorter-named APIs for creating types (Remove typefunc vars #324, Use VarId where appropriate #325, Jan 1st refactor #327)
• use VarId instead of MonoId in PolyData and gather_free_vars (Use VarId where appropriate #325)
• simplify ct_eval ... somehow (Remove typefunc vars #324, Dont rewrite nodes in ct_eval #328, Jan 8 refactor #330)
• remove debug data from typechecker nodes (Jan 1st refactor #327)
• don't do AST rewriting in the typechecker -- specifically in ct_eval (Dont rewrite nodes in ct_eval #328)