Check LHS expression of assignment

[klutzy]

This patch fixes rustc to emit explicit error if LHS of assignment is
not allowed.

Fixes #7507
Fixes #7508

[alexcrichton]

I have a feeling that this code doesn't belong in resolve, but rather somewhere in the middle of typechecking. I'm not entirely sure where it belongs in that code, however, although @nikomatsakis may have a better idea than I do.

[adrientetar]

Does it mean that we can't do tuple alloc as in Python?
That'd be cool to have as a feature.

[nikomatsakis]

Definitely not resolve. typeck seems like a suitable place (there are various options, but resolve is strictly for name resolution). It can re-use the existing routine ty::expr_is_lval.

[klutzy]

They are now in typeck/check/mod.rs.
Note: expr_is_lval thinks None is not lvalue, so None = Some(1) now produces different error message. (was: "cannot assign to immutable static item")

@adridu59 (a, b) = ... is unimplemented now, so this patch is just to produce a proper error rather than ICE.
There is some discussion at #7507, and we could open a RFC issue for the cool feature.