Add preliminary support for ghc-modi

[int3]

This enables the check and info functions to use ghc-modi if it is available. This addresses part of #48.

I had sketched an implementation for make as well, but it fails a lot of ghcmod-vim's tests due to DanielG/ghc-mod#275, so it's been left out for now.

Also, ghc-modi does not support spaces in filenames (DanielG/ghc-mod#295), so check and info inherit this issue.

The g:ghcmod_should_use_ghc_modi flag determines whether the plugin attempts to use ghc-modi.

[darthdeus]

[RubenAstudillo]

I'm using this, I really like it.

[ibotty]

is anything holding it back?

[denibertovic]

What's the status on this?

[gilligan]

@eagletmt can you merge this ? it seems to be working just fine and is a massive improvement in usability.

[rgrinberg]

+1

[denibertovic]

FYI, both issues mentioned by the @int3 in his first comment have been solved.

[ches]

It would be nice if :GhcModKillModi were mentioned in the doc file, and probably some general note about how the ghc-modi support works so that users have some troubleshooting hint to identify when an misbehaving process is to blame for things not working.

But otherwise 👍

[parsonsmatt]

I've been using this fork myself and it's a massive improvement. Anything we can do to bring it into master?

[liskin]

I rebased this on top of current master and changed a few things. Most notably I unified handling of newlines/boundary and added support for multiple ghc-modi instances (one per package). It works for me with a rather large multipackage haskell codebase that uses stack.

See #102 for my pull request.