Add functionality to ignore selfloops

[tlarock]

This PR adds the ability to ignore selfloops in path data to the Paths.add_path instance function and Paths.read_file class function. It also removes some redundant code from paths.py by changing read_file to use add_path with the expand_subpaths option.

I implemented this because I have a dataset that includes (apparently) meaningless self-loops that I want to be able to remove/include in my pipeline without preprocessing the data manually. The two main changes are:

1. Add remove_selfloops option to add_path with functionality that collapses consecutively repeated symbols into a single symbol. For example, the sequence ('a', 'a', 'b', 'b', 'c') will collapse to just ('a', 'b', 'c').
   • Note that since a sanity check in add_path is already looping through the elements of the path to ensure the separator character is safe to use, I only add comparisons and list appends to the loop so the change should not meaningfully impact computational complexity.
2. Change Paths.read_file to call cls.add_path rather than reproducing the same functionality. Moves expand_subpaths functionality to add_path rather than calling it on the whole object at the end.

In working with this code, I couldn't understand why Paths.read_edges is static and Paths.read_file is class. I think it makes sense for them to be consistent and I think they should both be static, since they both populate and return a new Paths object. I can change the decorator in this PR if desired.

Let me know if you have questions/comments/changes!

Edit: Meant to include a note that this is completely separate from my other open PR (#47).