RMSD calculation may lead to NaN values

[yzjin]

The RMSD collective variable implementation is problematic. It leads to numerical difference in the test case used here and NaN values even when the weights are removed here. The root cause of such behavior is unknown yet and require further investigation.

[pabloferz]

Our RMSD implementation now uses by default Kabsch algorithm as implemented in #321. Is this still happening or can we close the issue?

[yzjin]

I see! Closing this issue.

[pabloferz]

It's still may be worth making sure that it works against those test cases. Could you double check @yzjin?