Slow execution times for interp_hybrid_to_pressure with extrapolation

[kafitzgerald]

I've been digging into a report of slow execution times for interp_hybrid_to_pressure with extrapolation.

It looks like the for loops in the helper function _vertical_remap_extrap for interp_hybrid_to_pressure are particularly problematic in this case causing large dask task graphs to be generated and slowing execution times significantly.

I still need to refactor the code for the extrapolation of "temperature" and "geopotential" variables, but have some working code for "other" variables that's passing tests.