Fast disc emissivities

[fjebaker]

This is in part also to track how we want to expose the corona to disc part of Gradus modelling to the end user.

If we can get it fast enough, we can trace the disc emissivity functions directly when fitting (maybe using automated caching or similar) instead of pre-computing more tables.

[fjebaker]

The current time for the lamp post model is of order 100-200 milliseconds, which is too slow to be part of the fitting calculations.