Adds netCDF3 vs netCDF4 distinction to _automatically_determine_filetype.

[norlandrhagen]

Curious on what you think of this @TomNicholas.

• Allows _automatically_determine_filetype to guess between netCDF3 vs netcdf4.
• Adds a netCDF4 dependency to tests. 🤷

[codecov]

Codecov Report

Attention: Patch coverage is 89.47368% with 2 lines in your changes are missing coverage. Please review.

Project coverage is 88.14%. Comparing base (5eb15bb) to head (a2e2cd5).

Files	Patch %	Lines
virtualizarr/kerchunk.py	80.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #43      +/-   ##
==========================================
+ Coverage   87.66%   88.14%   +0.48%     
==========================================
  Files          13       13              
  Lines         851      869      +18     
==========================================
+ Hits          746      766      +20     
+ Misses        105      103       -2     

Flag	Coverage Δ
unittests	88.14% <89.47%> (+0.48%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

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

[TomNicholas]

This is cool! Presumably this has to open the file to read this data_model information? Is that likely to incur any significant cost over the opening that kerchunk already needs to do?

[norlandrhagen]

I think the read done by the NetCDF4 library is only reading attributes and not loading the entire file. Not 100% sure on this though.

[rabernat]

You can do this without opening the file with netCDF at all. Both file types have a "magic" at the beginning of the file.

def guess_file_type(fp) -> FileType:
    magic = fp.read(4)
    fp.seek(0)
    if magic[:3] == b"CDF":
        return FileType.netcdf3
    elif magic == b"\x89HDF":
        return FileType.hdf5
    else:
        raise ValueError(f"Unknown file type - magic {magic}")