Release 0.0.7

Some highlights over the ~ 700 commits

• The neighborlist is not static anymore. All data is screened in a two pass approach: first using pairs of residue-COMs + residue radii and then with actual atom-to-atom distances. This means that the old parameters interface_cutoff_Ang and nlist_cutoff_Ang are gone, and the user doesn't need to worry about "likely contacts" to speed up the computaton. Theres a number of new methods for this that can be used independently in the contacts and the `COM submodules.
• Parallel computing gets a working progressbar-like report
• Methods for making new ContactGroup out of existing ones, either by selecting:
  • by time frames
  • by residue (name or index)
  • by trajectory

allowing for finer-grained comparisons, e.g. between trajectories or between the beginning and the end the same trajectory.

• Access to KLIFS database can be done via UniProt Accession Code, kinase ID or structure ID
• New time-dependent plotting method ContactGroup.plot_timedep_ctcs_matrix, which allows for representations like this.
• The default cutoff is now 4 Å.

Some of the above changes were long overdue TODOs, many other changes, like smaller improvements (API calls have been simplified and harmonized across related methods, still some way to go) are spread throughout the 700 (!) commits.