Monkeypatch core.logic to make sets work

[arrdem]

This changeset tweaks the driver implementation to use Var.setRoot in a try/finally to establish an execution scope with a patched implementation of tree-term? that won't allow infinite recursion on sets.

Criticisms of this patch:

1. Blatant monkey patching
2. Assumes that kibit is only ever used via the driver interface, never as a library.

More of an effort could have been made to keep the test impact minimal, but that can be cleaned up.

Fixes #78

[danielcompton]

While I don't love monkeypatching as a general rule, no-one (including me) has provided a better option here, and not handling sets has caused so much trouble for people that I doubt that Kibit is all that useful as a library currently. I'll wait a few days before merging in case there are some good reasons not to merge.

[arrdem]

I'm not really a fan of monkey patching either, but if we're gonna monkey patch this is the nicest way we could possibly do it because it doesn't leak our patched predicate. I would rather see someone take a deep dive into core.logic and figure out what's going on here / how to support sets "properly", but I can't volunteer that kind of time.

[arrdem]

Posted a new diff. Removed irrelevant diff to the test suite, factored out a with-monkeypatches macro which may be more generally useful in the future.

[arrdem]

@danielcompton can we get a release whenever you get around to it?

[danielcompton]

lein-kibit 0.1.3 released. Thanks for your help!

[arrdem]

@danielcompton It's a janky workaround, but this may be worth a clojure@ patch ANN since it fixes a long standing bug.