- Removed toy_plot_helpers.py

- Checked all .py files
- Command used: flake8 --ignore=N806,N802,N803,E731

- nosetests for s-shooting simulator and analysis missing.

- Moved toy PES and integrator here from OPS fork.
- Set up basic structure for testing.
- Tests for toy PES done.

- Brownian walker seems to produce (almost) identical results as in the
  paper.
- Had to implement the OverdampedLangevinIntegrator.
- Bias function is also supported now.

- Per-snapshot and global time correlation function available.
- Had to add NonCanonicalSequentialEnsemble/Mover to loop over
  snapshots.
- Last step: how to add a user-defined bias function of a CV?

- Should change structure to use per-snapshot quantities similar to
  reactive flux and committor analysis.
- Simplified simulator (just use SequentialEnsemble).

- Improved Jupyter notebook.

- C program from C. Dellago to compare speed.
- Investigated performance problems: how does the number of frames
  influence speed (for a given trajectory length).
- Results in "check-scaling".
- D. Swenson knows about the problem (see
  https://github.com/openpathsampling/openpathsampling/pull/686 ).

- 1D example as in S-Shooting paper (double well potential).
- Last commit before ESDW-2.

- Maybe the ensembles should be changed to allow also reverse paths B->A
  (more efficient sampling).
- Forgot to (force) add the IPython notebook in last commit, here it is.