Dataset row id reuse old value when `enable_stable_row_ids` enabled

Description

Hi team, when enable_stable_row_ids enabled for a dataset, without checking out to a new branch, I expect all row ids to be monotonically increasing; but after restore, I found old row ids are reused.

Steps to reproduce

import tempfile
from pathlib import Path

import lance
import pyarrow as pa


with tempfile.TemporaryDirectory() as tmp:
    uri = Path(tmp) / "dataset.lance"

    ds = lance.write_dataset(
        pa.table({"value": [0]}), uri, enable_stable_row_ids=True
    )
    assert ds.has_stable_row_ids is True

    ds = lance.write_dataset(pa.table({"value": [1]}), ds, mode="append")
    before = ds.scanner(with_row_id=True).to_table()
    print(f"Version {ds.version} before restore:", before.to_pydict())

    ds = ds.checkout_version(1)
    ds.restore()
    print(f"Version {ds.version} after restore:", ds.to_table().to_pydict())

    ds = lance.write_dataset(pa.table({"value": [100]}), ds, mode="append")
    after = ds.scanner(with_row_id=True).to_table()
    print(f"Version {ds.version} after append:", after.to_pydict())

    assert before.to_pydict() == {"value": [0, 1], "_rowid": [0, 1]}
    assert after.to_pydict() == {"value": [0, 100], "_rowid": [0, 1]} # this is incorrect

Output

Before restore: {'value': [0, 1], '_rowid': [0, 1]}
After restore: {'value': [0]}
After restore and append: {'value': [0, 100], '_rowid': [0, 1]}

Expected behavior

For row ids after restore and append, the row ids should be [0, 2]

Lance version

ede0c330992571050f79b141f447a6254cd70970

Language binding

Python

Environment

No response

Logs / traceback

Start by running the Python reproduction using enable_stable_row_ids, checkout_version, restore, and append. Trace how row IDs are allocated across restore and the subsequent write, then add a regression test showing that the final IDs are [0, 2] rather than [0, 1].