datafacts: first-class typed parents field on DatasetMetadata (follow-up to #35)

Description

The pitch.

Follow-up #1 to #35 (closed as a duplicate of #31), the typed-lineage-field piece @dhve flagged as a clean, independent improvement.

## What
Adds a first-class, statically typed provenance field to `DatasetMetadata`:

```python
parents: list[DataFactsUrl] = Field(default_factory=lambda: list[DataFactsUrl]())
```

…and updates the merged `cid_facts` plugin (#31) to read it, keeping the old dict spelling working:

```python
def parents_of(dataset: DatasetMetadata) -> list[DataFactsUrl]:
    if dataset.parents:                       # typed field wins
        return list(dataset.parents)
    raw = dataset.metadata.get("parents", []) # legacy fallback, backward compatible
    ...
```

## Why
#31 declares lineage via `metadata["parents"]` — a free-form, untyped, undiscoverable dict entry. A typed field is statically checkable, IDE-discoverable, and self-documenting. Both @Hiten0305l and I independently proposed it on our closed PRs, which @dhve noted is a strong signal it's the right design.

## Non-breaking by construction
- `content_hash()` is **untouched**. Because `publish()` still folds parents into the content hash via `metadata`, a dataset declared through the typed field and one declared through the dict **content-address to the identical URL** — verified by a new test.
- #31's entire 44-test `cid_facts` suite passes unchanged.

## Tests
`packages/nest-plugins-reference/tests/test_cid_facts_parents_field.py` (6 tests): default empty, typed field read, legacy dict still works, address parity between the two spellings, typed-takes-precedence, and phantom-parent rejection via the typed field.

## Verify
```
uv run ruff check . && uv run ruff format --check . && uv run pyright && uv run pytest -q
# 740 passed, 1 skipped
```

## Coordination
@dhve — on #39 you asked @Hiten0305l and me to coordinate on this so it isn't submitted twice. You listed it as item #1 on my #35, and it originated in my closed PR, so I'm carrying it here. Happy to adjust if you'd prefer otherwise.