Account for inconsistently formatted `restart` and `mesh_mask` netcdf files.

[ma595]

We've noticed a number of consistency issues with grid, restart and mesh_mask files.

How should we proceed? We could provide a function to attempt to manipulate files into right format. Number of cases is growing.

DINO inputs

Grid

Dims: (depth, time_counter, y, x)
Coords: (time_counter, depth, nav_lat(y, x), nav_lon(y, x) )

Restart : CASE 1

Dims: (depth, time_counter, y, x)
Coords: (time_counter, depth)
Vars: (nav_lat(y, x), (nav_lon(y, x))

Solution: Promote nav_lat and nav_lon to coords.

Mesh mask : CASE 1

Dims: (depth, time_counter, y, x)
Coords: (time_counter, depth)
Vars: (nav_lat(y, x), (nav_lon(y, x))

Solution: Promote nav_lat and nav_lon to coords.

Gen data

Restart : CASE 2

Dims: (depth, time_counter, y, x)
Coords: (depth, time_counter, lon, lat)

Solution: Rename coords lon and lat to nav_lon and nav_lat.

Mesh mask : CASE 3 (but similar to 2)

Dims: (depth, time_counter, y, x)
Coords: (depth, time_counter, y, x)

Solution: Rename coords to nav_lon and nav_lat

Restart ref : CASE 3 (but similar to 2)

Dims: (depth, time_counter, y, x)
Coords: (depth, time_counter, y, x)