[WIP] Analysis tools for identifying ligand binding modes from MD or BLUES simulations.

[nathanmlim]

This PR adds in an analysis module I've been working on that automatically identify metastable ligand binding modes using tools from PyEMMA. Included is a juypter notebook that provides a brief walk through usage of the analysis modules, located in notebooks/example-id_bmodes.ipynb

The analysis modules can be accessed like blues.analysis import msm,cluster. Helper functions for these modules are located in tools.py

• msm submodule contains the ConstructMSM class:

ConstructMSM provides convenience functions for using PyEMMA.
This class assists in selecting the appropriate lagtime
through plots, TICA-transforms the input feature coordinates for
k-means clustering to discretize the trajectories, and constructs the MSM
for the purpose of identifying the metastable binding modes.

• cluster submodule contains the FindBindingModes class:

FindBindingModes provides functions to analyze and determine the number of
metastable modes from the resulting MSM data out of ConstructMSM() via
spectral clustering from PCCA.

• population submodule contains the BindingModeOccupancy class:

    BindingModeOccupancy provides functionality to calculate the occupancy
    of each defined binding mode, determined from FindBindingModes().

[nathanmlim]

This will need some changes with upcoming changes to our NCMC reporter.