Proposal: shake geometry handler for difficult geometry convergence cases

[yakutovicha]

Add random XYZ 5-10 * 10^-2 A displacement for the atoms usually helps to destroy the symmetry and forces gaussian not to take it into account. It often helps with geometry convergence.

[eimrek]

my preference would be to implement things like this in a RobustGaussianSCF workchain and not try to bloat the plugin too much, as it makes the plugin harder to maintain, for potential users to read/understand the code, and to prevent possible unexpected steps to run. This workchain could be part of this repo though. But on the other hand, in principle, as a feature that is disabled by default, this could also fit directly in the plugin.

[eimrek]

now that i reconsider, in the baseworkchain it would fit well 👍

[cpignedoli]

One should not forget to pay attention to constraints in case of shaking