depth.py missing Vstretching=5 for Norkyst_v3

[MTrodahl]

roppy was updated some years back to also include vstretching=5, used by Norkyst_v3 (both hindcast and operational model). S have a different expression, found in stretch5 in depth.py. Can get large deviations in depth levels at some locations for Norkyst_v3 if used with the former roppy version. Perhaps switch to the newest depth.py if s-layer files are used?

[knutfrode]

Hi Marta, and thank you for the report.

I was not aware of this as I have only used NorKyst3-files that are already regridded to z-levels, and in fact I am almost never using the ROMS native reader myself.

I copied now the newest version of depth from the Roppy package, as a helper module for the ROMS native reader.
Ideally, we should use the Roppy package as an external dependency, but @kthyng has made some adaptations for OpenDrift to account for zeta != 0, following this discussion from some time ago: #1235
Thus, after updating depth from Roppy, I now manually merged in those adaptations.
4c4ef7b
All tests are passing, and thus I believe/hope everything is ok.

I do not have a ROMS native file with Vstretching=5 readily available, so feedback is welcome if you encounter any problems (or success).