Segfault when doing trajin of certain single trajectory frames followed by another trajectory.

[drroe]

To reproduce:

parm tz2.parm7
trajin tz2.rst7
trajin tz2.nc
trajout seg_stripped.nc
go
quit

Confirmed if the first trajectory is Amber restart, NetCDF restart, NetCDF traj, TRR. Does not happen with mol2 or DCD so far.

[drroe]

The problem is when a trajectory with velocities and or forces is followed by a trajectory without velocities/forces. The output trajectory is initially set up to expect velocities/forces, so for the subsequent trajectory without that info the pointer is null. The reason the opposite way works is (I think) that each Traj_XXX class is set up to ignore forces/velocities in the incoming frame if the initial trajectory was not set up with them.

3 possible solutions:

1. Modify SetupEnsembleOut()/SetupTrajout() and their parallel equivalents (called from CpptrajState) to check for changed CoordinateInfo. Will require passing in the Frame(s) for possible modification.
2. Create a new function that does this check/modification, called from similar spots in CpptrajState.
3. Modify each Traj_XXX class to protect against this.

I'm leaning towards 2 or 3. 2 will make it so that warnings are easier and future trajectory classes will automatically be protected. 3 is good because they should be protected for both cases. Maybe I need to do a combination of the two, even though that's a lot of work.

[drroe]

Should be addressed by #1188