Fix function compute_scan_angle (QC flags), remove unused argument - #1780

Have not tested it, needs to rebuild ioda

needs to rebuild ioda

@chengdang actually for these you don't technically need to rebuild, just copy the new def_jedi_utils.py into the $JEDI_BUILD directory,, the path is in I think lib/python/....

thanks for getting this in and for giving it the attention it needed

@BenjaminRuston
Thanks! That's good to know. I was wondering if I need to re-build. Let me run some tests then.

ok yes the hack is to obviously get on the branch and then you have to copy the def_jedi_utils.py file to:

${JEDI_BUILD}/lib/python3.11/pyiodaconv

so that's a shortcut to do a quick test

Thank you @BenjaminRuston
I did a test with the aws file, here is the post-ioda view angle:

I think it is working as expected, but more tests with other sensors are probably needed.

ok @rajichidamb picked up your other changes and yes got some failures, switching to double to see if the failures change as this may fix those and cause other errors now elsewhere

think we have a situation where some converters used float and other double for the missing value

The np.full_like(..., float_missing_value) initialization resolves the uninitialized-memory problem in the old np.empty_like path, and the new QC handling fixes the old if not qc_flag[0] bug that silently discarded the QC array whenever the first ob was good. A few things before this merges:

1. Two call sites still use the old signature (silent breakage). src/hdf5/aws_2ioda.py (~L69) and src/satpy/irs_2ioda.py (~L310) still call with the old 3-positional form. Since the removed argument was first, Python won't error, the args just shift (zenith → altitude, altitude → zenith, zenith → qc_flag), so nearly all obs get flagged bad and sensorViewAngle comes out missing. These need updating in this PR (if there's not corresponding PRs?) Worth also double-checking cloudsat_netcdf_to_ioda.py, ozone_product_2ioda.py, wsfm_2ioda.py, and seviri_l1b_to_ioda.py, which import the function.

2. Altitude units are inconsistent across callers. The function divides sensor_altitude by 1000 (expects meters), but amsr2 (strip('km')), gmi (407.0), and tropics (550.) pass km (maybe others??): for those, the computed scan angle collapses to ~ the zenith angle. Pre-existing, but since this PR fixes the function, could we document the expected unit in the docstring and audit callers (or handle it here)?

3. Minor:

• dtype=float makes the output float64; suggest np.float32 for consistency with ioda_float_type (or just be rigorous about using the correct float type that's already defined)?
• double_missing_value is added but unused, was there a plan here?
• Reference/CI data may need regenerating for converters that pass QC flags (cowvr especially), since outputs legitimately change.

thanks for the re-review @BenjaminTJohnson , have made the changes and confirmed satellite altitudes for those calling this function at least are in meters

Reference files need update to match the new calculated sensorViewAngle