Adds halo argument to constructor to RegularCartesianGrid

[glwagner]

Plus some cleanup of input validation, and a test.

[ali-ramadhan]

Might be nice for the user not to worry about setting halo sizes but we can worry about doing this more generally/automatically once we have other advection schemes.

[codecov]

Codecov Report

Merging #508 into master will decrease coverage by <.01%.
The diff coverage is 78.57%.

@@            Coverage Diff             @@
##           master     #508      +/-   ##
==========================================
- Coverage   72.33%   72.32%   -0.01%     
==========================================
  Files          39       39              
  Lines        1717     1720       +3     
==========================================
+ Hits         1242     1244       +2     
- Misses        475      476       +1

Impacted Files	Coverage Δ
src/grids.jl	79.24% <78.57%> (-0.76%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 525d5c0...c2b4c1f. Read the comment docs.

[glwagner]

Quite agree, though we will still need a constructor that allows us to set halos.

The design I am thinking of is that every term in the equations is associated with a function; something like

required_halo_size(::BiharmonicDiffusivity) = (2, 2, 2)

Then in the Model constructor (or Equation constructor, see #259) we can determine the required halo size for the grid based on the maximum required halo size for each term.

Each grid then needs a function

with_halo(halo, grid)

which returns a new grid that is identical to the old except for the addition of halos.