Migration lint for use crate::... has incorrect suggestions for braced use

[alexcrichton]

The following code:

#![warn(rust_2018_breakage)]
#![feature(rust_2018_preview)]

use foo::{a, b};

mod foo {
    crate fn a() {}
    crate fn b() {}
}

fn main() {
    a();
    b();
}

yields:

warning: Absolute paths must start with `self`, `super`, `crate`, or an external crate name in the 2018 edition
 --> src/main.rs:4:11
  |
4 | use foo::{a, b};
  |           ^ help: use `crate`: `crate::a`
  |
note: lint level defined here
 --> src/main.rs:1:9
  |
1 | #![warn(rust_2018_breakage)]
  |         ^^^^^^^^^^^^^^^^^^
  = note: #[warn(absolute_path_starting_with_module)] implied by #[warn(rust_2018_breakage)]
  = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in the 2018 edition!
  = note: for more information, see issue TBD

warning: Absolute paths must start with `self`, `super`, `crate`, or an external crate name in the 2018 edition
 --> src/main.rs:4:14
  |
4 | use foo::{a, b};
  |              ^ help: use `crate`: `crate::b`
  |
  = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in the 2018 edition!
  = note: for more information, see issue TBD

but when rustfix automatically applies the suggestions they end up breaking compilation because the suggested fix isn't syntactically valid

[nikomatsakis]

I'll have to figure out what we can do here. I suspect handling this in its full generality will be hard but that if we're careful we can address most cases (e.g., the one here) — or at least detect when we are giving wrong suggestions. Also relevant to my comment in 50672.