Add CSV I/O for PointDataset (cluster-friendly alternative to JSON)

[Jammy2211]

Motivation

Strong-lens cluster workflows can involve tens or hundreds of multiply-imaged
background sources, each currently stored as its own PointDataset JSON file.
Editing hundreds of JSON files by hand (to add/remove images, correct
positions, retune noise values, paste in flux/time-delay measurements) is
fiddly and error-prone.

A single CSV — one row per image, grouped by name — is a much more natural
cluster input. It opens cleanly in a spreadsheet, diffs line-by-line in git,
and can be generated by standard astronomy tooling. JSON stays the canonical
exact-round-trip format; CSV becomes the hand-editable form preferred for
clusters.

Proposed schema (one row per image)

| name | y | x | positions_noise | flux | flux_noise | time_delay | time_delay_noise |

• name, y, x, positions_noise are always required.
• flux / flux_noise and time_delay / time_delay_noise are optional;
  columns dropped when blank for every row.
• Rows are grouped by name → one PointDataset per group.

Library changes (PyAutoLens)

1. autolens/point/dataset.py
   • PointDataset.to_csv(file_path) — csv stdlib only, no pandas dependency.
   • classmethod PointDataset.from_csv(file_path, name: Optional[str] = None);
     name= required when the CSV has multiple groups, auto-picked when there
     is exactly one.
   • Module-level output_to_csv(datasets: List[PointDataset], file_path) and
     list_from_csv(file_path) -> List[PointDataset].
   • Fix the stale PointDict reference in the module docstring — the class
     no longer exists; the workflow is now lists of PointDataset.
2. autolens/__init__.py: re-export output_to_csv and list_from_csv
   alongside the existing output_to_json / from_json re-exports.
3. test_autolens/point/test_dataset.py: round-trip tests covering
   • positions-only
   • positions + fluxes + fluxes_noise_map
   • positions + time_delays + time_delays_noise_map
   • positions + fluxes + time_delays
   • heterogeneous list (one with fluxes, one without) — verifies
     optional-column handling at the list level

Workspace changes (autolens_workspace) — follow-up PR

1. scripts/point_source/simulator.py: after every al.output_to_json(...)
   that writes a PointDataset, emit an equivalent al.output_to_csv(...)
   with the matching .csv filename and add narration — JSON is the exact
   round-trip, CSV is the hand-editable spreadsheet form preferred for
   clusters.
2. New scripts/point_source/features/csv.py:
   • write a synthetic list of ~3 PointDataset objects to a single CSV via
     al.output_to_csv
   • reload with al.list_from_csv
   • build [FitPointDataset, ...] per loaded dataset and print info
3. Do not change the modeling scripts yet; JSON remains the canonical
   modeling input until the cluster examples land.

Non-goals (explicitly deferred)

• FitPointDataset.to_csv residual / chi-squared export — stretch goal;
  follow-up issue if worth it.
• Introducing a PointDatasetList class — keep list-level helpers as plain
  functions for now.
• Adding pandas as a dependency.

Classification

Library-first. Workspace examples ship in a follow-up once the library PR is
released.