transpile: Handle module name collisions

[randomPoison]

Note that the current code is directly from Codex, I haven't yet reviewed and cleaned things up and am putting this up mostly for visibility.

In testing against libgit2, we ran into a case where module names could collide, leading to a generated Rust file not being included in the build. The situation is that we have both a hash.c file and some files in a hash directory, e.g. hash/sha1.c. C2rust then wants to generate both

mod src {
    mod hash;
}

and

mod src {
    mod hash {
        mod sha1;
    }
}

In this case, the latter is what gets generated, effectively dropping src/hash.rs from the Rust build.

This PR has c2rust detect these collisions and rename the colliding module, so we'd generate something like

mod src {
    #[path = "hash.rs"]
    mod hash_;
    mod hash {
        mod sha1;
    }
}

[ahomescu]

The transpiler renamer uses the scheme {name}_{i}, it would make sense to keep that for consistency.

[ahomescu]

This is quadratic in the number of modules, it might be worth optimizing this a little. Maybe sort them first, or use a trie?

[ahomescu]

"{}_0" for consistency with the transpiler? Or anything other than a trailing _?

[ahomescu]

Doesn't insert_path_module already checks for uniqueness? Why do you need the extra trailing underscore?

[ahomescu]

Don't we also need to check the new module for collisions?

[thedataking]

nit: couldn't