FEAT: Precomputed calibration model
This MR adds a new Precomputed
calibration model. The idea of this is to use a set of precomputed calibration realisations and sample in the index of the line rather than the underlying model parameters.
This is similar to the method proposed in https://arxiv.org/abs/2009.10192 and has two advantages over the current method:
- only one new parameter per detector, rather than the standard 20 with the cubic spline.
- it can incorporate the physical calibration model described in the paper above.
I branched this from !1201 (merged) so the changes are a little confusing, but the new code, is essentially just the Precomputed
class.