SimStore: Support for OpenMM snapshots - #949

@dwhswenson I don't have any old netcdf available that I would like to load up (all of them used to take a long time to even open up).
If needed, I could spin up a small simulation in an old OPS version and see if I can load it with this branch. Just ping me if you want me to do that 😉

@sroet : The main thing I want to check is that any existing analysis (especially anything that requires snapshot coord/vels, like calculating a new CV on existing data or even simulation restarts) still works. So if you have any scripts along those lines that will work on a small dataset, then please do! (Obviously, some analyses, like TIS, only work with a lot of data.)

I've checked a few basic analysis scripts that I have, and it seemed to work. In principle, everything should work -- older files will just create snapshots with, e.g., snap.kinetics.engine is None. There might be some problems with loading such snapshots from netcdfplus and re-saving as SimStore, but switching storages is going to be hard anyway.

This is ready for review and comment. I will leave it open for at least 48 hours, merging no earlier than Tue 22 Dec 07:00 GMT (08:00 local).

@sroet : The main thing I want to check is that any existing analysis (especially anything that requires snapshot coord/vels, like calculating a new CV on existing data or even simulation restarts) still works. So if you have any scripts along those lines that will work on a small dataset, then please do! (Obviously, some analyses, like TIS, only work with a lot of data.)

The .nc file that I managed to run

• could load up an old trajectory (Think > 2.5 years old)
• load state definitions (from a different file, saved by OPS master), and use them for scheme.initial_conditions_from_trajectories(traj) and save these initial conditions on this branch
• load state definitions (from a different file, saved by this branch) and use them for scheme.initial_conditions_from_trajectories(traj) and save these initial conditions on this branch

This seems to me that at least continuing simulations (into new storgaes) is possible and is as far as I have time/hardware to spend on this.
(As getting the state stores to work was already a pain, as I could not get a working python2 env up and running to load the old marshall code back in)

Thanks @sroet : If this can load a trajectory that's that old, including recalculating things (as would be needed from initial_conditions_from_trajectories if you don't have CVs cached, which sounds impossible in your case), then this should be fine. Merging!