not completely) with Amber.

In particular, the 1-4 electrostatics now matches between Gromacs and Amber, but
the 1-4 L-J terms are *roughly* 2x larger, but not exactly 2x larger.  Not sure
what's going on here.

Add a warning if the [ pairs ] doesn't match the 1-4 list, and zero out 1-4 interactions not specified

- Add protection for instantiating an AmberParm from unsupported functional
  forms
- Do the same with ChamberParm
- Fix Travis failures with regards to NBFIXed systems in AmberParm or
  ChamberParm, since they were broken by moving NBFIX support into Structure.
  They should work correctly now.
- Change load_from_rawdata and load_from_structure to from_rawdata and
  from_structure, respectively (and deprecate, but do not eliminate, the former)
- Update Lee-Ping's script to use from_structure instead of load_from_structure

Both version 5 and version 4 use `mdrun` as the name of the executable. Version 5 is distinguished by having `gmx` whereas version 4 does not. Switched order of 'if' statements to catch this.