Fix segmentation faults when reading in trajectories without forces/velocities after reading in trajectories with forces/velocities

[drroe]

Version 7.10.0. Should address #1187.

In the case where loading in a trajectory with velocity/force information before a trajectory without velocity/force information, the output trajectory would be set up with that information expected. When encountering a trajectory without that information a segfault would occur. This PR fixes the behavior so that zeroes for forces/velocities are written instead of segfaulting when this occurs. This is fixed for both trajin and ensemble in serial and in parallel.

Adds a check that replica indices map to the same values when loading in multiple ensembles; otherwise sorting may not work properly.

Adds 3 new math functions: hyperbolic sin, cos, and tan.

Adds a new keyword for ensemble, bytemp to force sorting by replica temperature (instead of replica indices).

Adds more tests and updates the documentation.

[drroe]

According to git bisect, the first bad and last good commits:

commit 066c404e430670a427abe7d694e02f88802a673b (HEAD)
Author: Daniel R. Roe <daniel.r.roe@gmail.com>
Date:   Tue Jul 14 16:12:44 2026 -0400

    Version 7.10.0. Minor version bump for fixing handling remd ensembles with changing vel/frc info. Also added 3 math funcitons.

commit 784e9fd9fd0aa6a0c3abaa48bd8739082a86cf2a
Author: Daniel R. Roe <daniel.r.roe@gmail.com>
Date:   Tue Jul 14 15:40:43 2026 -0400

    Hide debug info

[drroe]

Sometimes I am a complete dingdong...

Need to use NcTest() to compare NetCDF files, not DoTest() 🤦‍♂️ . The version bump was causing the files to look different.

[drroe]

The Jenkins failure is a non-specific one for the CUDA build. Potentially the CUDA runner is having issues. Local CUDA compiles and tests fine so I am going to merge.