Further improvements of tide interpolation performance

[veenstrajelmer]

• Tide interpolation: from extrapolated from rasterized #730
• replace GTSM4.1preliminary with GTSMv4.1 in example script
• Add non-p-drive source for EOT20 tidal model #837
• clean up and simplify code, cognitive complexity is 18 instead of max 15 (tpxo removal will fix it)
• consider removal of tpxo, added in: Add tpxo to tidemodels #365, also update modelbuilder example notebook. Then also update modelbuilder notebook
• add translated components like eps2 and La2 to translate testcase
• add link to delft3d-tide manual in docstring as source for convention component names. Are all names correct?
• maybe interpolation on 0-360 grid is faster for models like FES
• for 0to360 model like fes, also add point just before or after 0 meridian
• check if interp is faster on separate datasets instead of concatenated ones
• enable github testing (plifile is local, only tpxo/gtsm is on opendap)
• consider download of gtsm/eot instead of online connection, performance for large files is not great, but downloading also takes time+space
• consider EOT20 via opendap instead of P (SL/MHD): https://openadb.dgfi.tum.de/archive/eot20/ocean_tides
• alternatively use fsspec to open file from url:

file_nc_git = 'https://github.com/Deltares/dfm_tools/raw/test-large-files/tests/GTSMv4.1_TS_2014_M2_rasterized.nc'
import fsspec # requires aiohttp for HTTP
# more info: https://github.com/Unidata/netcdf4-python/issues/1043#issuecomment-697416049
f = fsspec.open(file_nc_git)
ds_git = xr.open_dataset(f.open(),engine='h5netcdf')