Full precession simulations should be optimisied and may have a bug #160
Description
Describe the bug
Related to the discussion in #159
Equation (8) of the paper at: https://doi.org/10.1107/S2052520619007534 describes how to integrate for some rocking curve I_tbgiven the length of the g vector, the precession angle and the unprecessed excitation error.
Our current code goes via a route that finds the average excitation error (which isn't strictly needed), it may be wrong and probably a lot slower than evaluating the above integral directly.
However, I'm fairly sure that for large precession angles (here something like g*phi > 2 * max_excitation_error) the contribution of the rocking curve isn't that important and using the two beam case (ie. approximate_precession = True) we provide should be fine. Even in the low integration case the two beam approximation is probably better than (say) linear drop off