Move to Chrome's fork of Hunspell

[anaisbetts]

This PR moves Spellchecker to use the Chromium fork of Hunspell, which supports a format invented by Chromium called "bdict" files. bdict files are smaller and less weird (i.e. no aff files), and we're also guaranteed that Chrome keeps them up-to-date.

Unfortuately, compiling with bdict support disables the built-in support for dic/aff files, so this is a breaking change (though I've tried my best to keep the legacy API compatible).

The breaking change is, setDictionary now expects a Buffer as its second parameter, which is a loaded bdic file:

spellChecker.setDictionary('en_US', fs.readFileSync('path/to/en-us.bdic'));

A new method getURLForHunspellDictionary now will return the URL to download the dictionary file, given a language code.

TODO:

• Windows
• OS X
• Linux

[anaisbetts]

This is working okay in my testing, but still has some weird behavior where the first word is marked as an error if it's capitalized.

[Thomas101]

I've given this a go on osx and it looks like it works! Would be great to get this merged in and released as it looks like it closes #49 and #10

[ghost]

I'm also keen for this functionality, is there anything that's stopping us from merging this? Is there anything still left to do before we can merge this?

[anaisbetts]

@callodacity The capitalization bug is the only thing stopping us, my lldb skills are lacking, maybe @zcbenz can have a look if he's got a hot second? We're using this as a fork in electron-spellchecker and the Slack Desktop app and it mostly works (Chrome's dictionary has a few oddities, but in general it's good)