Fixes for membrane proteins

Resolves two issues found when simulating membrane proteins.

1. Lipids of the boundary 'evaporating'/dancing into the outside of the sphere An additional CL option is introduced '--rest_shell_width' to Qligfep which allows you to restrain solute atoms near the edge of the sphere with 10.0 kcal mol-1 A-2 sequence restraints. Simply give the width of the shell in Angstrom. A decent setting seems to be 3 A, default is 0 A.

2. Qprep assumes protein density for all solute atoms Qligfep now calculates the number of POP lipid carbon atoms in the sphere and adjust the density in the qprep input file. Lipid carbons are assumed to have the same volume as octane carbons, I'm not excluding head group carbons as this would be additional work but this could be added. Should help avoid extreme pressures and resulting artifacts, e.g. violent ejection of ligands into the void.

@jesperswillem I tested it with one some files from @bananatana and Hugo, but there might be some edge cases that break things, in particular with the quick and dirty pdb parsing in the code I added to functions.py file. Might want to check that. Also I couldn't really think of a good single char flag to set the restraints in the CLI but if you want to handle it differently it is probably best to change it now before people start putting this in their pipeline.

/Florian

qusers / qligfep

Fixes for membrane proteins #19