OptionT implements MonadCombine, but doesn't meet the laws

[mikejcurry]

Whilst looking at OptionT today and adding some tests in #692, I noticed that the OptionT has a MonadCombine instance, plus implementations for empty and combine.

The tests exercise MonadCombineTests.monad instead of MonadCombineTests.monadCombine, which means that the implementations of empty and combine are not tested against the laws (or anything).

I briefly looked at increasing the test to monadCombine, but the current implementation of OptionT does not pass the laws, at least not for OptionT[List, Int].

I noticed that #687 proposes to remove, at least temporarily, the MonadCombine for XorT - if a similar approach is warranted here I can take a look. By similiar in this instance, I mean that it might make sense to reduce the implementation to Monad until the laws pass.

[ceedubs]

Good catch! Thank you for bringing this to attention.

I think that the right approach would probably be to:

1. Remove this instance
2. Make sure we have an issue for separating the weak and strong laws suggested by those comments in the XorT code
3. Create new issues for adding back in the XorT and OptionT instances once the laws work has been finished.

How does that sound?

[mikejcurry]

That sounds sensible. I can put together a PR(after #692 - as both PRs will touch OptionTTests) and create issues for tracking.

[ceedubs]

I'm going to go ahead and close this out since it was resolved with #698 and #695 tracks the follow-up.