refactor: dedupe FootprintPlotter, import from cs_util

[cailmdaley]

Summary

FootprintPlotter was duplicated between sp_validation.plots and cs_util.plots, with the local copy stale relative to the canonical cs_util version. This PR deletes the local copy and imports cs_util's FootprintPlotter directly at every call site — the elegant end state, as if the class had always lived in cs_util.

What changed

• plots.py — the local FootprintPlotter class is gone; the module no longer references it at all (no re-export shim).
• cosmo_val.py — drops from .plots import FootprintPlotter; uses the module's existing cs_plots alias → cs_plots.FootprintPlotter().
• notebooks/cosmo_val/catalog_paper_plot/plot_gaia_sky.py and notebooks/demo_create_footprint_mask.py — import from cs_util.plots import FootprintPlotter directly.
• pyproject.toml — cs_util>=0.1.5 → >=0.1.9 (guarantees the import resolves).

⚠️ Intended breaking change

from sp_validation.plots import FootprintPlotter no longer resolves. This is deliberate — cs_util is the home for the class, and every caller in the repo now imports it from there. No shim is kept.

Caller audit

cs_util 0.1.9's FootprintPlotter is a superset of the old local API (same __init__(nside_coverage=32, nside_map=2048) and public methods, plus vmin/Milky-Way/extent improvements). All call sites were inspected and are compatible:

• cosmo_val.py:plot_footprints — FootprintPlotter() + create_hsp_map + plot_region
• plot_gaia_sky.py — FootprintPlotter(nside_coverage=32, nside_map=2048) + plot_area(..., vmax=...)
• demo_create_footprint_mask.py — FootprintPlotter() + plot_region / plot_footprint_as_hp

Tests now run in CI

Previously no unit tests ran in CI: the only live workflow (the docker build) did python -c "import sp_validation", and ci-build.yml was dead (triggered on master, Python 3.8, conda, setup.py test).

• tests/test_plots.py (new) — guards the dedup contract (cs_util is the home; sp_validation.plots/cosmo_val don't re-expose it) and exercises create_hsp_map on a tiny input.
• testpaths fixed from ["sp_validation"] (matched nothing under the src layout, so pytest recursed the whole tree and choked on a stray *test* script) → ["src/sp_validation/tests"].
• @pytest.mark.slow on the catalog-data-dependent tests (additive-bias reads the real SP_v1.4.5 catalog; catalog_paths_exist checks files on disk), restoring a fast hermetic -m "not slow" suite (2:54 → 7s); catalog_paths_exist also skips when the data isn't mounted.
• Dockerfile installs the [test] extra; deploy-image.yml runs pytest -m "not slow" inside the built image (which carries the full system-dependency stack), so tests run on every push.
• Deleted the dead ci-build.yml.

— Claude on behalf of Cail

[cailmdaley]

Live smoke test passed inside the shapepipe container (Python 3.12, cs_util 0.1.9):

from sp_validation.plots import FootprintPlotter
import cs_util.plots
fp = FootprintPlotter()
# FootprintPlotter.__module__ -> cs_util.plots
# FootprintPlotter is cs_util.plots.FootprintPlotter -> True
# list(fp._regions.keys()) -> ['NGC', 'SGC', 'fullsky']
# all('vmin' in r for r in fp._regions.values()) -> True

Re-export is transparent; same class object is returned, _regions matches cs_util's (with vmin). Ready for review.