Pp add warning and tests

[pp-mo]

Shadow to #145
Just to preview tests for that, since CI runs here !

[valeriupredoi]

nevermind, the test I was saying it fails locally is fine, I was using a dev Xarray where I'd mess things up - all fine for me locally:

==== 786 passed, 49 skipped, 109 warnings in 22.59s ======

This is with most of the deps installed from conda-forge - see attached env file. I really strongly recommend you folks use conda-forge deps and not PyPI ones, those are always a bit blergh from many causes. Do you want me to pop a conda env file and create the env from that?

condaenv_ncdata.txt

[valeriupredoi]

ah I can't push to branch, but here is aht I was just about to push: in the GHA workflow, pop a call to a minimal environment file:

        with:
          miniforge-version: latest
          activate-environment: testenv
          environment-file: environment.yml

and get dask (at least) from conda forge in a conda environment.yml:

---
name: ncdata
channels:
  - conda-forge
  - nodefaults

dependencies:
  - dask
  - netCDF4

I can help you get those test deps to be test deps in the pyproject toml file afterwards, but I reckon stuff like xarray and dask should definitely be installed in the main dev env and from conda-forge 🍺

[valeriupredoi]

nevermind my blabbering above - I had to go use your full test suite - that test is not from ncdata, it's from iris sample data testing, I can see it failing now, locally and on forked GHA, sorry for having been Thomas the Unbeliever 😆

[valeriupredoi]

Suggested change

	if not list(set(var.attrs.values()) & set(std_axes)):
	if not [s for s in std_axes if s in str(var.attrs.values())]:

[valeriupredoi]

this will fix that failing test, and all the almost 1000 tests pass now 😃 The problem is that this method (and the previous approach) are not 100% exhaustive in figuring out which are coords that should have realized data, so a few of these warnings will be raised for those that the check misses them, but beats me how I can create a bulletproof check, given the variety of names and metadata, and differences between Zarr2 and Zarr3 (alas, most of them don't raise the warning)

[pp-mo]

NB As explained valeriupredoi#2 I actually think it's better to go back to #145.
Please see my responses there ...