New "best cell" attribute for the Crystal model

[dagewa]

Following dials/dials#670 we would like a "best cell" attribute on the Crystal model, which can be set by post-integration refinement by a program like dials.two_theta_refine. It will be useful to keep this separate from the Crystal's existing static A matrix, because that will have been used for spot prediction, whereas in the context of dials.two_theta_refine, the "best cell" will not have been refined jointly with crystal orientations U and therefore should not be used for spot prediction.

I'm not completely happy with the name best_cell (is it really "best"?) and would welcome any alternatives.

[dagewa]

I'm looking at this now and don't think we need any new methods or public attributes in Crystal. Rather, accessors can have an extra parameter to choose between the usual cell used for prediction and a postrefined=True cell. This second option would not have been used for prediction for integration, but may be determined by some post-integration refinement algorithm, like dials.two_theta_refine. The parameter name postrefined seems more appropriate to me than best.

[dagewa]

fixed by b9d6bb2