Add support for bucket expression to table scans

### Feature Request / Improvement

For time partitioning, we can express time range expressions and they lead to partition pruning when planning an Arrow scan:

```py
scan = table.scan(
row_filter=And(
GreaterThanOrEqual("event_ts", start),
LessThan("event_ts", end),
)
)

arrow_table = scan.to_arrow()
```

It would be useful to be able to filter by other hidden partitioning transforms, such as buckets -- e.g. filtering on `bucket[16](user_id) in {0, 1, 2, 3}`, and getting partition pruning whenever possible based on the underlying table partitioning specs (falling back to filtering rows when files cannot be pruned).

Start at the table.scan and scan.to_arrow entry points described in the issue, then trace the existing time-range expression handling used for Arrow partition pruning. Check how partition specs and row filters are applied; done means bucket expressions such as bucket[16](user_id) can prune files when possible and fall back to row filtering when they cannot.

Written by the indexing model from the issue text.