Replaces deepcopy with a smarter copy implementation, which only copies the instance attributes.

This is a speedup of about 20x (and yes, the deepcopy was getting annoyingly slow in some of my code).

Still using `partial_fill` when creating FileTreeQuery

Added test for I/O to both JSON and pickle files

This is now the new default behaviour when loading a new file

This change is not backwards compatible!

There were some subtle bugs in the use of get_all, which should now be fixed. They were caused by having multiple optional variables and the result not being deterministic if these were iterated in different orders.

more sense, and to make it easier to un-ambiguouosly identify the location
of a Match in a filetree hierarchy

actually exist on disk (and corrsepond to files).

Includes now a tutorial on the FileTree format

reverting back

needed for python 3.5