GeezerVison

This adds a function to "copy" pointmatch dictionaries such that changes to the match arrays of the original do not affect the copy.

As background for why this is useful, BigFetA has use cases where it transforms pointmatches in-place. This becomes a problem when doing multiple-step in-memory alignment or trying to interrogate residuals. To be on the safe side I've been copying the matches before each step, which is very expensive on large alignments.

The test I added uses the same match objects (via copy/paste) as the integration test, but I am surprised to see that the format of the dictionaries doesn't match what I was seeing in real pointmatch collections, e.g.

"p": [[1, 2, 3], [1, 2, 3]] (pointmatch collection)
vs
"p": [[1, 1], [2, 2], [3, 3]] (test fixture)

This actually doesn't affect the way I have the copy implemented, but I'm not sure if this is a long-standing typo or something that changed without us noticing.