Improve support for NetCDF fill value and valid range attributes

[axelboc]

Part of #1918

A while back, we added basic support for NetCDF's _FillValue attribute.

In this PR, I go a step further by implementing the spec more accurately, and by supporting three additional NetCDF attributes: valid_min, valid_max and valid_range. These attributes are documented here: https://docs.unidata.ucar.edu/netcdf-c/current/attribute_conventions.html — the section on the valid_range attribute actually talks about the order in which the attributes should be used. It also explains that _FillValue acts as an (invalid) upper or lower bound depending on if it's positive or negative.

So in this PR:

• I add support for valid_min, valid_max and valid_range.
• I add support for valid_min/max/range and _FillValue to the NX Line and NX Heatmap visualizations.
• I make sure _FillValue acts as an invalid upper or lower bound (instead of just as an invalid exact value).
• I bring the floating point equality check for _FillValue in line with the spec (i.e. difference lower than two epsilons).
• I add mock datasets and tests.
• I create a new netcdf visualization pack for the new code.

[axelboc]

/approve

[loichuder]

Fell swoop 👏