chore: stop tracking simulator-regenerated datasets

[Jammy2211]

Summary

• Untrack ~220 files under dataset/ that are regenerated by simulators in scripts/{imaging,interferometer,multi,group,cluster,point_source,weak}/simulator.py and friends. They were perpetually dirty in git status (143 staged + many unstaged).
• They were already covered by the dataset/ rule in .gitignore, but had been committed before the rule existed, so git kept tracking them.
• Real observational data and simulator inputs are kept tracked (40 files total):
  • dataset/imaging/cosmos_web_ring/ — JWST COSMOS-Web; loaded by scripts/imaging/start_here.py, scripts/multi/start_here.py
  • dataset/imaging/slacs1430+4105/ — HST SLACS lens; used by scripts/guides/plot/examples/
  • dataset/group/102021990_NEG650312660474055399/ — real catalog object; loaded by scripts/group/start_here.py
  • dataset/interferometer/uv_wavelengths/sma.fits — synthetic SMA uv-coverage; read by scripts/multi/features/imaging_and_interferometer/simulator.py and several modeling/aggregator scripts
  • dataset/imaging/los_halos/los_halo_list.npy + los_sheet_values.npy — line-of-sight halo data; read by scripts/imaging/features/advanced/los_halos/simulator.py
  • dataset/.gitignore — the inner * + !.gitignore marker

.gitignore change

Replaces:

dataset/

with:

dataset/**
!dataset/.gitignore
!dataset/imaging/cosmos_web_ring/**
!dataset/imaging/slacs1430+4105/**
!dataset/group/102021990_NEG650312660474055399/**
!dataset/interferometer/uv_wavelengths/**
!dataset/imaging/los_halos/los_halo_list.npy
!dataset/imaging/los_halos/los_sheet_values.npy

This documents which datasets are deliberately tracked. Adding new real data later: git add -f dataset/path/to/data.fits and add a matching !dataset/... line to .gitignore. (The inner dataset/.gitignore still ignores everything by default, so -f is required for new files in keep-dirs.)

Effect

• No source-code changes. Working-tree files stay on disk.
• git status will go silent for the simulator-output dataset tree.
• Fresh clones still get all real-data dirs (cosmos_web_ring, slacs1430+4105, 102021990_NEG..., uv_wavelengths, los_halos npys) — start_here.py tutorials work out of the box.
• Fresh clones lack simulator-output dirs (e.g. dataset/imaging/simple/); running python scripts/imaging/simulator.py regenerates them.

Test plan

• CI green on the branch
• After merge: clone fresh, confirm dataset/imaging/cosmos_web_ring/, slacs1430+4105/, 102021990_NEG.../, uv_wavelengths/sma.fits, and los_halos/*.npy are all present
• Run python scripts/imaging/simulator.py, confirm dataset/imaging/simple/ regenerates and git status stays clean
• Run python scripts/imaging/start_here.py (loads cosmos_web_ring), confirm it still finds the data

🤖 Generated with Claude Code