Contact trajectories

[dwhswenson]

The main purpose of this is to introduce a ContactTrajectory subclass of ContactObject. It also comes with a few related additions and tools. Major step toward #53.

ContactTrajectory

This is essentially a list of single-frame ContactFrequency instances. A contact map for a specific frame can be obtained by either:

# these are the same
contacts = contact_trajectory[frame_num].residue_contacts
contacts = contact_trajectory.residue_contacts[frame_num]

The normal ContactTrajectory is a Sequence; there is also a MutableContactTrajectory which is a MutableSequence.

This PR will not include the movie making aspects suggested in #53 -- that will be left for future work.

ContactObject.from_contacts

For efficiency, the easiest way to calculate the contact trajectory is to have a loop that calculates the contacts for each frame, and then to create a ContactFrequency from atom_contacts and residue_contacts. We did not have a method to create a ContactFrequency from contacts before, so I added it. Like many things, it isn't allowed for ContactDifference, where it is overridden with a NotImplementedError.

The input to ContactObject.from_contacts can be either ContactCount instances, or the underlying collections.Counters (as well as, of course, the necessary information for ContactObject.__init__).

WindowedIterator and ContactTrajectory-specific implementation

The basic idea here is to facilitate rolling-average type behavior (perhaps in our case, rolling-frequency). By iterating over an underlying sequence, the idea is to return sequence indices that should be added to/subtracted from the rolling average. There's a general implementation (perhaps useful in other code!) and a specific use of it to get rolling contact frequencies.

---

Tasks

• Implement and test ContactObject.from_contacts
• Implement core of ContactTrajectory
• Test core of ContactTrajectory
• Test with sliced trajectory
• Add compatibility checks where relevant in ContactTrajectory, with tests
• Implement MutableContactTrajectory, with tests
• WindowedIterator and tests
• RollingContactFrequency and tests
• Docstrings
• Add example
• Add to docs

[dwhswenson]

Quick comments on the codeclimate complaints:

• from_contacts has too many arguments (3 times): True, but unavoidable. And I'll also make the arguments that, (1) most of those arguments typically use defaults; (2) while this is part of the API, it's aimed less at normal users than at developers. Both of these make the too-many-arguments thing less of a problem in my mind.
• file contact_maps.py is too long: Yes. I'm not going to change that in this PR, but I was already thinking about reorganizing some of the code here. That will be a separate PR.
• TODO: yes, I added a TODO... that shouldn't be blocking, though

I'll silence those issues for now: wontfix on the too many arguments, confirmed on the too long file. Approving the changes.

[dwhswenson]

This is ready for review. Once this one is in, I'll cut 0.6.0.

[sroet]

LGTM codewise, some doc changes and leftover debug code in the tests

[dwhswenson]

Made changes as suggested, @sroet. Thanks for the quick review!

[sroet]

LGTM, feel free to merge on green