Parmed

This is a small feature-enhancement release with one small fix
- `parmed.load_file` and `Structure.save` convenience methods now work with file-like objects that support seek and tell operations
- Copying or slicing a `Structure` preserves `TER` cards when writing to a PDB file
- Atom type names are taken from OpenMM's `Atom.id` attribute where applicable

This release has improved support for converting between different formats, including OpenMM, Amber, CHARMM, and GROMACS.

It has numerous bug fixes, and all users are encouraged to upgrade to this version.

This is a minor bugfix release:
- Fixes a nuisance parsing some PDB files with insertion codes and ANISOU records (762)

There are a number of improvements and bug fixes including:
- Improved atomic number determination for mol2 files (695)
- Fix OFF file writing for solvent boxes (702)
- Fix bug in storing angle types for CHARMM systems (704)
- Eliminate severe performance degradation of OpenMM simulations with the NetCDF reporter (722)
- Support loading molecules from the `rdkit` package
- Update for small changes to OpenMM API
- Support native visualization through the nglview backend
- Support reading SDF-format files (through rdkit)

And many other small improvements.

This is a major new stable release of ParmEd since version 1.1. An abbreviated list of changes and improvements are listed below:
- Unified object model built on `Structure` that facilitates file format translations and integration with OpenMM
- Support for reading and writing GROMACS topology and GRO files
- Automatic file loading based on file contents (rather than filename extension)
- Read many new file formats, including Mol2, GROMACS topology, GRO, PDBx/mmCIF, Amber OFF libraries, Amber parameter and leaprc files, CHARMM residue templates, OpenMM XML files, and more
- Expressive selection syntax for slicing substructures and views of `Structure`
- Flexible molecule splitting functionality within `Structure`
- Support Python 2 and Python 3 in the same code base through the `six` module
- Overhauled the coordinate and unit cell handling functionality, creating a unified interface for all objects based on the n-dimensional array in `numpy`
- C++-accelerated Amber topology file reader
- Abstract file reader capable of reading compressed and remote files (and remote compressed files)
- Convenient `Structure.save` method that converts between different file formats automatically
- Added `gromber` action to ParmEd interpreter to facilitate converting from GROMACS to AMBER file formats
- Added ability to instantiate parametrized `Structure` from OpenMM Topology and System objects
- Added robust unit handling and dimensional analysis through the `unit` package that is part of OpenMM
- New package organization -- there is now a single subpackage `parmed` that contains all of the modules and packages. `chemistry` has been renamed `parmed` and `ParmedTools` has been moved to `parmed.tools`
- Renamed scripts from `parmed.py` and `xparmed.py` to `parmed` and `xparmed`.
- New, extensive API documentation generated automatically by Sphinx
- Many bug fixes and stability improvements
- Added serialization support (so that the objects in ParmEd can now be pickled)

This release brings some substantial changes and API stability. The major changes here that will affect users upgrading from older versions are
1. Python 2.7 or later is now required. Support for earlier versions of Python has been discontinued.
2. The `numpy` package is now required and is used for coordinate and unit cell handling
3. OpenMM support requires version 6.3 or later of the OpenMM library
4. The following packages have been renamed:
- `chemistry` -> `parmed`
- `ParmedTools` -> `parmed.tools`
- `ParmedTools.ParmedActions` -> `parmed.tools.actions`

Making the above substitutions should be sufficient to fix existing scripts or programs to work with the newest version of ParmEd.