Add FsEventWatcher Support. fix #7

[andelf]

This is ispired by the original work of https://github.com/octplane/fsevent-rust .

Currently you can only call .watch() once, and .unwatch() left unimplemented.
Good news is that Drop works. :)

UPDATE: .watch() .unwatch() works.

r: #12 #7
@octplane

[passcod]

I'm going to leave this to @octplane to evaluate and decide if it's mergeable or if some of it should be split back to fsevent-rust.

I'd also like to have watch and unwatch working completely before merging, unless y'all think that's too much work.

[andelf]

:) done.

[andelf]

@octplane if some of this should be split back to fsevent-rust?

[octplane]

LGTM. I will probably later remove some code from here to avoid duplication with my binding, but is seems reasonable to merge as-is if it is working correctly :). Great job @andelf!

[octplane]

@andelf: yes ! your comment arrived while I was writing mine :)

[andelf]

@octplane @passcod
This PR can be merged and @octplane and I will fix the duplicated code betweent fsevent-rust and rsnotify. ;)
The API will be compatiable.

BTW, what's the suitable behavior of after unwatch() all paths in a Watcher? Stop the Watcher or triger an error?

[passcod]

Leave the Watcher open for further watch() calls. If it's best to spin down the inner implementation or somesuch, do that (and spin it back up on a .watch()), but as long as it's not Dropped I would expect to be able to use it.

[passcod]

I'll probably merge in either a few hours, or tomorrow. Thanks both for the great work!

[passcod]

That should say new_fsevent().

[passcod]

Comments in the code are minor style nitpicks, ignore them; I'll fix in the merge.

[passcod]

This should probably be a Result<(), String>