Fix QuantileDigestIterator bugs from #115

[jonhehir]

The QuantileDigestIterator claims to perform a post-order traversal, but it actually has been performing in-order traversal. This is troublesome for most Q-Digest algorithms, which depend on post-order traversal.

The upperBoundQuantile() and lowerBoundQuantile() functions were also not correct and have been removed and replaced with a simple cumulativeProportion(). (To obtain lower bound and upper bound quantiles correctly, you need to separately perform forward and reverse iteration.)

These are breaking changes, but the functions have no known usage.

See #115 for details and example.

[tdcmeehan]

Thanks a lot for fixing these bugs. I'll have time to review this properly in the couple of weeks. Thank you for writing test cases that demonstrate your fix.

[jonhehir]

Thanks, @tdcmeehan! Do you happen to know if this is being used anywhere? I want to make sure I'm not breaking anything, but I don't see any usage in OSS Presto.

[tdcmeehan]

I don't think so. IIRC, I had an idea to generate approximate histograms using Q-Digest using this iterator, but I don't think I ever wrote the function that would use it.

[jonhehir]

Thanks! In essence that's what I was looking to do here too, but with empirical distribution functions instead of histograms. Since the postOrderTraversal function is private, I wanted to use QuantileDigestIterator instead.

[jonhehir]

@tdcmeehan: Any chance you can review this soon? Thank you!

[tdcmeehan]

Thanks @jonhehir