Use cf-conventions to infer dimension names

[SeanBryan51]

Currently ILAMB requires dimensions for time, latitude and longitude in NetCDF files to be called time, lat and lon respectively which is not CF-compliant. This change allows ILAMB to be more flexible in inferring the spatiotemporal dimension names by examining standard CF attributes on coordinate types.

[SeanBryan51]

Hi @nocollier, requesting your review for this pull request

[nocollier]

Thanks for your contribution and your patience while I get to looking at it.

I think this is great, and appreciate your effort to make ILAMB better. I just have a few comments to make in general in case you find yourself staring at a lot of my (ugly) code:

• I want to be careful about CF-conventions / CMOR requirements. Your contribution does not apply here as it makes ILAMB more flexible than it was, but in general I try to require very little from model output. In my experience, it is work for model centers to convert output and if I can just be more flexible, then great.
• My apologies for that code, it was written by ~8 years ago me staring at CMIP5 output which was not as well formatted as CMIP6 and before packages like xarray had really matured.
• I have a a ilamb3 rewrite underway, but it has been a lot of work to try to cleanup everything and re-implement it. I will think about porting your functions there too.

I am going to accept and merge this, and then will run some larger tests on our bigger comparisons in the week to come.

Thanks!

[SeanBryan51]

Hi @nocollier, can you push a new release to conda/pip containing these changes? We require these changes for our work and it would be convenient if they were available directly via conda.

@abhaasgoyal