Refactor the um2nc submodel

[atteggiani]

Refactor the commands.um2nc module::

• [ ] Split functions into multiple "one-task" functions, for simpler unit testing.
  • Split in pre & post processing funcs for each data library? (split mule & iris ops out to separate functions)
• [ ] Use built-in constraints arg in iris.load() to filter to required cubes first and reduces repeated filtering further down in the function body
• [ ] Refactor I/O to separate functions
• [ ] Refactor include/exclude functionality:
  • Return list of names to keep
  • Validation: ensure no common items between include/exclude
  • Simplify logic so default args are empty containers?
• [ ] Fix exception based control flow in um2nc.cubewrite()
  • Manually check for ancilliary file (other file types? make check explicit for code clarity)
  • Split modification & write steps into separate funcs
• [ ] Saving: can cubes & global metadata be modified, then save() relying on the file extension?
  • Can this get around I/O sprinkled throughout main()?

[truth-quark]

Additional tasks:

• [ ] Add pytest dependency
• [ ] Add coverage.py & pytest-cov dependencies
• [ ] Add coverage.py configuration