Horizontal wind components expected by CCPP schemes and their standard names

[grantfirl]

During the course of trying to arrive at a defensible and cohesive set of standard names, the question has often come up (and continues to; see, e.g. NOAA-EMC/fv3atm#337): what should the horizontal components of wind be called? Our working hypothesis with existing UFS physics is that all of the schemes merely expect that the two horizontal wind components are orthogonal to each other and are tangential to the local surface. The CF standard names include both [x,y]_wind and [eastward,northward]_wind. We originally chose to use [x,y]_wind since it implies orthogonality and is arguably more general than calling it [eastward,northward]_wind. For most physics scheme types, this should be "good enough". An issue could arise for gravity wave drag schemes (whose orography statistics may assume true zonal/meridional directionality) or even surface layer schemes (who could potentially adopt direction-specific roughness lengths?) however. Those schemes should be able to request true zonal/meridional winds if they need them, which would either require an interstitial physics scheme to rotate the wind components (which @dudhia says happens with WRF) or the host model to have already done such a rotation (which I believe happens with FV3 before physics is called and hence why only [x,y]_wind are sufficient for UFS physics today).

Some questions:

1. Should we require that CCPP schemes use true zonal/meridional winds and not simply orthogonal components?
2. Should we carry two sets of horizontal wind variables [x,y] and [zonal,meridional]? For FV3, these would be identical (I think!), but may not be for other hosts. If both existed, one would also need to carry directionality information about the [x,y] winds in case the host supplies only those and yet, say, a GWD scheme expects zonal/meridional, so that a proper rotation can be done in an interstitial scheme.

@gold2718 @SMoorthi-emc @dudhia @mdtoy @ligiabernardet @cacraigucar