Add snapshot_id parameter for Datafusion IcebergTableProvider (Distributed reads)

### Is your feature request related to a problem or challenge?

Currently table scans are pinned to current snapshot for catalog backed providers
`
Ok(Arc::new(IcebergTableScan::new(
table,
None, // Always use current snapshot for catalog-backed provider
self.schema.clone(),
projection,
filters,
limit,
)))
`
This is fine for single-node reads, but for a distributed engine, each remote node must read against the same snapshot_id the coordinator planned against. Otherwise a concurrent write during reads can result in worker nodes reading incorrect versions of the catalog.

related to https://github.com/apache/iceberg-rust/pull/2613

### Describe the solution you'd like

add a passable snapshot_id to IcebergTableScan to pin reads

### Willingness to contribute

yes

Start with the IcebergTableProvider scan construction shown in the issue and trace IcebergTableScan, then review related PR #2613. Determine how a planned snapshot_id reaches distributed reads; done means callers can pass it and remote workers read the same snapshot instead of the current catalog state.

Written by the indexing model from the issue text.