Vendor eftest and fix #26

[camsaul]

Fixes #26

While fixing this I had to do so much icky monkey patching that I decided the "right" thing to do here would be to just vendor Eftest entirely since we've diverged so much at this point. This let me clean up some existing monkey patching as well.

I renamed eftest.* to mb.eftest.* to prevent issues where both versions are present and tried to delete some stuff we're clearly not using (e.g. Eftest's version of the Junit output or test-var discovery code) but there's a lot more cleanup and consolidation we could do in the future.

More on why I decided to vendor this:

To fix #26 I had to monkey patch eftest.runner/test-vars, a 31-line function that calls a ton of private functions, with basically a line-for-line replacement that looked exactly the same but bound hawk.parallel/*parallel?*... so vendoring made the fix for #26 a one-line change instead of almost 40 once I copied the function over and did the alter-var-root nonsense.

This wasn't the first time we did something like this either, for example in

hawk/src/mb/hawk/parallel.clj

Lines 15 to 17 in a8f3e8a

	(def ^:private synchronized? (complement parallel?))
	(alter-var-root #'eftest.runner/synchronized? (constantly synchronized?))

we did something similar, which I refactored out in this PR

Overall vendoring lets us eliminate a lot of the monkey patching or duplicate code and should make future changes much easier to do in the future since we can modify the underlying Eftest code without monkey patching and other hacks

[crisptrutski]

🤝

[technomancy]

This looks good but IMO it could use some description of why we decided to vendor this; what our eftest fork does that the upstream doesn't.