Add "intersect" method to GenomicArray

[etal]

Just collect the results of "iter_ranges" into a single DataFrame and return a new GenomicArray.

This was an odd oversight, but apparently CNVkit has been using iter_ranges instead to do what it needs.

[etal]

For speed: Collect the matched row indices and do a single slice operation on the original dataframe -- should be much faster than the current approach of extracting each row into a tuple and collecting the tuples into a new dataframe. (See #346)

That won't work on its own if mode='trim' is specified. In that case, iter_ranges provides a simple solution. It may still be significantly faster to do the index-only approach as with mode='inner', and then separately collect just the bins that need to be trimmed, generate them separately, and concatenate the two dataframes.