Subset global field with arbitrary (rectangular) lat / lon corners

[jpolton]

Story:
Lazy load in a global sea surface temperature data set with the fancy COPERNICUS read method.
Cut out domain of interest with input bounding coords.
Tricky bit: deal with wrapped coordinates

lat = [50,60], lon=[-10,5] # UK
lat = [50,60], lon = [350,5] # UK
lat = [50,60], lon = [5,350] # zonal band excluding the UK

Deal with dateline:
lat = [-10,10], lon = [170, -170] # somewhere in the Pacific

Steve's work address much of the wrapped coordinates issue, but the implementation was incomplete

AC:

• subset + plot on battery of example subdomains for COPERNICUS
• subset zos across longitude join for SENEMO_1ts_20210101_20211231_grid_T.nc in /projectsa/COAsT/SENEMO at latitude=0
• subset zos across longitude join for SENEMO_1ts_20210101_20211231_grid_T.nc in /projectsa/COAsT/SENEMO at latitude=85
• unit_testing updated and works

[jpolton]

Ticket 327 is probably a good place to start. It briefly describes what Steve tried to do and the code he sought to replace.
It has an associated Pull request. Which I didn't like (it didn't work) and so created the following ticket.

Ticket 425 discussed a issue found in Steve's implementation that manifest in example_files/subsetting_by_lon_lat.py but was I think a result of his method coast.general_utils.subset_indices_lonlat_box() being to limited in its scope relative to the original code. (It appears that Steve enforces that the subset must be a square rather than a rectangle. Maybe there are other issues too)

The pull request is also missing Unit_testing so that is also missing from Steve work.

[lukegorman]

issue part of #425