Improve calibration for speed & precision

[kif]

The calibration relies on a least square fit of theoretic (from calibrant) vs calculated 2theta position with the one obtained from the geometry and the position of the peaks.

• For speed: when wavelength varies, both the calibrant and the geometry part need to be updated. Fitting in q-space (rather than 2theta) would simplify the problem with only one function to call and thus remove some of the python overhead
• For precision: the density of ieee754 floating point values is maximal around 1.0 thus optimal results should be obtained with delta-q around 1 ... this can be used to discriminate between the different units possible.