Add some assertions that are helpful for trouble shooting.

[Amplifier-seller]

1. Checking input CV files

If there got a problem in the input CV files, these commits allow rid-kit to transmit the raw error messages produced by PLUMED to users for debug. Rid-kit provides little corresponding information in the workflow log file and only records a "CUDA error #1" in the gmx log. The check.py called by the newly added assertion parses the CV files with PLUMED in advance to ensure their validity and facilitating trouble-shooting.

2. Assertions about trust_lvl_1 and std_threshold

Too large trust_lvl_1 and too small std_threshold may emit all the data points from the training set and interrupt the workflow. When this happens, these assertions allow rid-kit to inform users to adjust the parameters.