[Bug]: `nodata`/`_FillValue` value is not being set in output products

[scottstanie]

Checked for duplicates

Yes - I've already checked

Describe the bug

When plotting the static layers, I noticed that the empty region was 0s instead of nan like it is for the CSLCs.

What did you expect?

I expected

1. the nodata to be nan like the CSLC, or
2. to have it marked in the metadata

even if we change (1) to match, we should still mark it in all output files

Reproducible steps

(mapping) [staniewi@aurora gslcs]$ rio info --nodata NETCDF:t071_151204_iw1_20190704.h5:/science/SENTINEL1/CSLC/grids/VV

(mapping) [staniewi@aurora gslcs]$ rio info --nodata NETCDF:static_layers_t071_151204_iw1.h5:/science/SENTINEL1/CSLC/grids/static_layers/heading

[scottstanie]

Adding the docs on missing data for CF Conventions: http://cfconventions.org/cf-conventions/cf-conventions.html#missing-data

• it could be done by adding a fillvalue argument required for init_geocoded_ds

• COMPASS/src/compass/utils/h5_helpers.py

  Line 104 in 5205eca

  shape=shape)

  would then have fillvalue=fillvalue for the require_dataset
• if we're not using h5netcdf we'll have to manually set `.attrs['_FillValue'] = fillvalue

[vbrancat]

This should be addressed by #167