ENH: Always calculate norm factor.

The Clease calculator can now also calculate CF from scratch.

There is an issue with numerical stability with cutoffs which are extremely close to an internal distance, e.g. if the max cutoff diameter is equal to the lattice vector. Some clusters which are expected to be found, are suddenly missing.

The main point here, is that I believe the normalization factors should just always be calculated, rather than inserted after a Clease calculator is created.

This MR contains a few "controversial" changes, so I'm pinging @davidkleiven if you have a bit of time?