Add ability to use a custom field_table file

[spencerkclark]

This adds the ability to add a custom field_table in the configuration dictionary to override the inferred version based on the microphysics scheme used. The behavior is implemented as described in this design document that @mcgibbon outlined earlier after a discussion with me and @oliverwm1.

I definitely see value in automatically setting the field_table used to prevent user-error as much as possible, so eventually I think we could look to somehow support the SHiELD use-case from that perspective; however, that introduces a few more complications that we'd have to think through:

• The microphysics scheme used is set with different namelist parameters than imp_physics.
• The new boundary layer scheme requires a new tracer for TKE, though one can still run with older PBL schemes, which do not require this additional tracer (so the field_table not only is dependent on the microphysics scheme, but also is dependent on the boundary layer scheme).

I don't think there's a particular downside to allowing users (at their own risk) to explicitly set the field_table used, so hopefully we're all on board for the spirit of this change!

Thanks @mcgibbon for pointing me to the tempfile module; that was quite useful for writing these tests.

[spencerkclark]

Many thanks for the review @mcgibbon -- all good comments. I'll push some updates tomorrow morning.