Added spatial index to OverlayNG PolygonBuilder.placeFreeHoles

[arriopolis]

This PR adds a spatial filter to the polygonbuilder that is part of overlayng. This was already suggested in a comment present in the previous version.

I did some performance measurements too, to see how much time it saves. For the tests, I computed the difference between two MultiPolygons, one being a region and the other being lakes and rivers inside that region, both taken from OpenStreetMaps. Here are the performance results:

Region	Old time (ms)	New time (ms)	Data set
Zuid-Holland	194	189	zuid-holland.zip
Netherlands	902	903	netherlands.zip
Spain	1286	1168	spain.zip
Germany	2774	2163	germany.zip
France	4489	3322	*

*The data set for France is too big to upload, it's available upon request.

For the measurements, I took the average of five runs. I also checked that the areas of the resulting shapes match exactly in both runs.

[dr-jts]

Looks like a good improvement. Is there a particular reason you chose to use HPRTree - e.g. performance or memory usage? (It does seem like a good choice - just curious if you compared it to say STRtree).

[arriopolis]

Thanks!

I did think a bit about which one to use, and I found that there was not much of a difference timing-wise.
My hunch is that with this improvement, this part of the computation becomes only a very small fraction of the total runtime of the multipolygon difference operation that I was testing it on, so it makes sense I couldn't tell them apart from the timing results.
Then I found these two lines in the HPRtree.java which eventually made me decide to go with HPRtree:

* NOTE: Based on performance testing, 
* the HPRtree is somewhat faster than the STRtree.

I have no further experience with either, so I guess from my perspective I have no logical argument to favor the one over the other.

[dr-jts]

Yes, I'm pretty sure HPRtree is more space efficient than STRtree, and has comparable performance. It should also be a good choice for very small index sizes (in this case it skips the sort step, and uses simple linear search).