Client: monomorphic `let rec`

See merge request !49

More consistent handling of errors in Solver, Infer and FTypeChecker, with source locations

See merge request !48

The problem with 'range' is that this word has a much more general
meaning, it could be an integer range; typically it is also used in
comments to describe range of levels or scopes.

In the context of Menhir, which is centered on parsing, 'range'
clearly associates with 'source range'. But for more general project
this association is lost and 'range' is confusing.

More efficient typechecking of System F pattern-matching

See merge request !46

# Conflicts:
#   client/test/TestF.ml

More informative error messages when typechecking System F variants

See merge request !47

Add a notion of type equality in the client

See merge request !44

These comments should appear in F.mli too, but I'm waiting for feedbacks before I copy-paste them.

The term Refl is a proof of a type equality TyEq(ty1,ty2) and Use is a construct that introduce a type equality in the typing context.

This will be useful when we translate System F terms to graphs (a "structure" node of the graph will be represented by a Structure.structure).