Ensure truly safe loading of netcdf drivers.

[SpacemanPaul]

Reason for this pull request

Distracting messages when reading from non-netcdf files when netCDF4 is not installed as per #2282

Proposed changes

• Ensure netCDF driver can be loaded enough to cleanly report that it cannot be loaded enough to actually use

• Prevent datacube.driver.netcdf.__init__ from importing stuff that might break.

• Closes #xxxx

• Tests added / passed

• Pull Request Title will make sense in ODC Release Notes

---

📚 Documentation preview 📚: https://opendatacube--2305.org.readthedocs.build/en/2305/

[codecov]

Codecov Report

❌ Patch coverage is 75.00000% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.12%. Comparing base (9ccc81b) to head (e5a3f8d).

Files with missing lines	Patch %	Lines
datacube/drivers/netcdf/driver.py	62.50%	3 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2305      +/-   ##
===========================================
- Coverage    86.13%   86.12%   -0.02%     
===========================================
  Files          155      154       -1     
  Lines        16955    16949       -6     
===========================================
- Hits         14605    14598       -7     
- Misses        2350     2351       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pjonsson]

I'm not sure how things are wired up, but isn't there a missing if driver_init is None: before this try?

It's possible I'm in the wrong place, but there should be a check for the new None value returned from reader_driver_init and writer_driver_init somewhere.

[pjonsson]

Please add the missing None check where it should be before merging this.

Besides that I think this looks great, can't wait to not have those warnings in the logs!

Edit: I've spent a fair amount of time on debugging ODC-related things now and I find it difficult to read and navigate both git commit logs and git blame when the same logical change is spread out over multiple commits. The scattered commits also frequently have a commit message that mostly makes sense when read in sequence of the commit stack of the PR, not when looking at an individual commit. For this particular PR, I'm guessing it should consist of one or at most two commits.