Reader for netCDF look-up tables

[adamcpovey]

This isn't exhaustively tested but, since people are using it, I feel we should start working on the code together.

Outstanding issues:

• The dimension ordering in the tables is different to that in ORAC. As we believe the table ordering is optimal, the LUT interpolation code should eventually be updated to reshape the arrays. Then, the copying code and temp arrays in read_sad_lut_nc() can be removed.
• The Python scripts aren't designed to deal with the many and varied LUT names that Don is currently producing. If we're going to use inhomogenous sets going forward (e.g. different version numbers across sensors), we will need to reconsider how phase is input into the scripts.
• The various volcanic ash types should probably be given a class. I propose renaming Eyj to Ash.

[simonrp84]

Ready for final review, @adamcpovey and @andyprata

[simonrp84]

Any final comments before I merge this, @andyprata?

[simonrp84]

Changes now made and merged. Thanks for your work on this, everyone!