MIR-borrowck: diagnostics for closures are much worse compared to AST-borrowck

[pnkfelix]

When compiling with MIR-borrowck via -Z borrowck-mir, the diagnostics printed when errors involves borrows/moves into the bodies of closures (where the origin path is from the outside of that closure) are pretty bad.

For example, for the test src/test/compile-fail/borrowck/borrowck-closures-two-mut.rs, the AST-borrowck error messages tend to highlight the particular uses within the closure body, while the MIR-borrowck error messages just treat the whole closure as a black box.

Specifically, look at this AST-borrowck error:

error[E0499]: cannot borrow `x` as mutable more than once at a time (Ast)
  --> ../src/test/compile-fail/borrowck/borrowck-closures-two-mut.rs:55:24
   |
54 |     let c1 = to_fn_mut(|| set(&mut *x.f));
   |                        --           - previous borrow occurs due to use of `x` in closure
   |                        |
   |                        first mutable borrow occurs here
55 |     let c2 = to_fn_mut(|| set(&mut *x.f));
   |                        ^^           - borrow occurs due to use of `x` in closure
   |                        |
   |                        second mutable borrow occurs here
56 |     //~^ ERROR cannot borrow `x` as mutable more than once
57 | }
   | - first borrow ends here

and compare that to the corresponding MIR-borrowck error:

error[E0499]: cannot borrow `x` as mutable more than once at a time (Mir)
  --> ../src/test/compile-fail/borrowck/borrowck-closures-two-mut.rs:55:24
   |
54 |     let c1 = to_fn_mut(|| set(&mut *x.f));
   |                        ----------------- first mutable borrow occurs here
55 |     let c2 = to_fn_mut(|| set(&mut *x.f));
   |                        ^^^^^^^^^^^^^^^^^ second mutable borrow occurs here
56 |     //~^ ERROR cannot borrow `x` as mutable more than once
57 | }
   | - first borrow ends here

What MIR-borrowck needs, if we can somehow get it, is the equivalent to the "previous borrow occurs due to use of x in closure" and "borrow occurs due to use of x in closure" messages, with the corresponding (very) precise spans.