PPS cloud water path composites fall through to the default enhancement

[djhoese]

The Problem

While working on the "warning for default enhancement" in #3237, Claude identified a likely bug. Here's Claude's summary:

satpy/etc/composites/visir.yaml defines two parallel sets of NWCSAF/PPS cloud water path
composites, and only one of them has a matching enhancement.

Set A — cloud_water_path, ice_water_path, liquid_water_path,
which set standard_name: cloud_water_path / ice_water_path / liquid_water_path.
There is no enhancement section for any of those three standard names in
satpy/etc/enhancements/generic.yaml.

Set B — pps_cwp, pps_iwp, pps_lwp,
which set standard_name: pps_cwp / pps_iwp / pps_lwp and do have
matching colorize enhancements
using cmic_cwp_pal / cmic_iwp_pal / cmic_lwp_pal.

So loading a Set A composite matches no enhancement section and falls through to the default
wildcard section, which applies a dynamic linear stretch. The GEO equivalents are fine — they have
cloud_liquid_water_path / cloud_ice_water_path enhancements.

The tests point at Set A

test_nwcsaf_comps
parametrizes cmic_cwp_pps, cmic_lwp_pps and cmic_iwp_pps onto the Set A names:

"cmic_cwp_pps": ("pps", "cmic_cwp", None, "cmic_cwp_pal", None, "cloud_water_path", ...),
"cmic_lwp_pps": ("pps", "cmic_lwp", None, "cmic_lwp_pal", None, "liquid_water_path", ...),
"cmic_iwp_pps": ("pps", "cmic_iwp", None, "cmic_iwp_pal", None, "ice_water_path", ...),

This goes unnoticed because the test's assertions are
guarded by if flavour == "geo":,
so the pps cases load the composite and enhance it but never check the result.

Notes

• Set A uses ColorizeCompositor, which has been
  deprecated since Satpy 0.39
  in favour of ColormapCompositor. It also colorizes inside the compositor, so a linear stretch is
  being applied on top of already-colorized output.
• Surfaced by the new default-enhancement warning in Add warning to default enhancement #3237, which reports
  No YAML enhancement found for 'cloud_water_path' (and the other two) during the test suite.

Possible resolutions

1. Deprecate/remove the Set A composites in favour of Set B, and repoint the test parametrization.
2. Keep Set A and add enhancement sections for its three standard names.
3. Either way, drop or extend the if flavour == "geo": guard so the PPS cases actually assert
   something.

Addition by Dave: If the colorize/colormap compositor is meant to pre-enhance the dataset then it would seem best to have an enhancement defined that is the static limit of the enhanced data (0-1?). Otherwise the dynamic stretching of the default will always change the result.

[djhoese]

Oh even worse, the ColorizeCompositor says to use the ColormapCompositor but that is also actually deprecated and says to do it "manually" with a single band compositor and an enhancement definition.