Potential bug in the reconstruction of MPAS fields from edges to cell centers

The reconstruction from edges to cell centers as written in MPAS-Framework assumes that the values at all boundary edges are zero. There are two MPAS-O fields that use this reconstruction, `sshGradient` and `surfaceVelocity` (and their constituents `sshGradientZonal`, `sshGradientMeridional`, `velocityZonal`, and `velocityMeridional`).

The zero normalVelocity boundary condition at edges is desirable. **It is less clear that zero ssh gradient boundary condition is desirable, particularly for sea ice dynamics.**

I have run idealized MPAS-Ocean standalone tests to expose the issue, which are detailed in a comment below and show a reduction in the amplitude of the reconstructed field on the order of 10-50% at the boundary cells, with some spatial variability that is likely related to the orientation of the current relative to boundary edges.

MALI also calls the MPAS-Framework routine `mpas_reconstruct` but I am unsure about whether the zero-valued Dirichlet boundary condition is desirable for those fields.