CDK-988. Add a long range partitioner with fixed size bounds.

[rdblue]

@tomwhite, I'd like to do a release soon. Is this something you want to get in?

[tomwhite]

Thanks for the feedback @laserson and @rdblue! I have updated the PR to use the start value for each bucket in the range.

I've also addressed the previous feedback with a couple of exceptions:

• I left the range size as a long. Most of the time it will be an int, but I don't see why we need to limit it to be an int. (For example, you could imagine very sparse data with large range sizes.)
• The implementation of projectStrict could be improved (for ranges), but what we have there now is good enough.

[rdblue]

One more note: any idea how to extend this for int, float, and double types? For the 4-byte types we could simply promote them and run the long/double partitioner?

I think we would need a separate partitioner for a double fixed range, which I think will be valuable for geo purposes since lat/long values are usually decimals. To do this we could just add a "type" option to the existing "range" JSON. Just want to hear your thoughts here so we have a plan for adding it later that doesn't conflict with this work.

[tomwhite]

@rdblue Thanks for all the feedback, and for the guidance on implementing the project methods. I've added tests for them now, and addressed all your other points.

Regarding double fixed size range - it should be straightforward to add it in the future with a "type" option as you suggest.

[rdblue]

A added a couple of minor comments, but this looks good overall.

[rdblue]

+1

Thanks, Tom!