Extensive component tests needed

[mcgibbon]

There are a lot of edge cases we currently don't test on CliMT components. We should create unit tests for as many of these as we can think of. Some examples:

GFS:

• With or without contained Prognostic objects
• When contained Prognostic objects have different output dims than itself
• When passed Prognostics attempt to diagnostically output quantities that GFS steps (like air temperature)
• When passed Prognostics step quantities GFS steps, and when they step quantities GFS does not step
• When multiple passed Prognostics compute tendencies for the same quantity
• In moist mode and dry mode
• With different numbers of vertical levels

I'll work on making these after the Sympl 0.4.0 updates. I may continue using this issue as a pad for examples of needed unit tests. Feel free to add any you think of, if you'd like.

[mcgibbon]

Some of these have been implemented in #68, but not all of them.

[mcgibbon]

Still need to go through every component and think about what options should be enabled/disabled for tests.