High order wind stress re-mapping

[kieranricardo]

Currently using first order conservative regridding to map the wind-stress from atmosphere grid to the ocean grid. With N96 + 0.25 degree ocean, the ocean is at a ~6x higher resolution, making these fluxes blocky on the ocean grid and affecting the curl of the wind stress of the ocean grid.

Wind stress will currently have an $O(\Delta x)$ error on the ocean grid, so the curl (calculated on the ocean grid) will have a $O(1)$ error (i.e. it's inconsistent). This is probably important even on a 1 degree ocean grid.

This can be fixed by using a higher order re-mapping.

[aekiss]

Here's what we did in ACCESS-OM2 COSIMA/access-om2#52