Add support for SetDivergence to DivergenceF2C

[dennisYatunin]

This PR fixes a bug in ClimaAtmos related to the advection of precipitation.

Advection is implemented using the DivergenceF2C operator, and, at the bottom cell center, we need to enforce a zero-divergence boundary condition for precipitation. We initially implemented this using SetDivergence, but this is not supported for DivergenceF2C and was ignored due to the following fallback method in finitedifference.jl:

boundary_width(::DivergenceF2C, ::AbstractBoundaryCondition) = 0

This PR adds support for SetDivergence to DivergenceF2C, as well as its operator matrix. It also includes a new unit test for these additions.

• Code follows the style guidelines OR N/A.
• Unit tests are included OR N/A.
• Code is exercised in an integration test OR N/A.
• Documentation has been added/updated OR N/A.

[dennisYatunin]

@simonbyrne Do you think it would be a good idea to get rid of the AbstractBoundaryCondition fallback methods for boundary_width in the future? They could potentially lead to more silent bugs like this if we keep them around.

[simonbyrne]

@simonbyrne Do you think it would be a good idea to get rid of the AbstractBoundaryCondition fallback methods for boundary_width in the future? They could potentially lead to more silent bugs like this if we keep them around.

What's the bug?

[simonbyrne]

Is it #1212?

[szy21]

No, it's here: This line doesn't give an error when SetDivergence is not defined. It just ignores the boundary condition.