[rkrishn7]

Adds a required GroupValues::find: for each input row, the stored group id or None, without inserting and without changing stored values or ids.

fn find(&mut self, cols: &[ArrayRef], groups: &mut Vec<Option<usize>>) -> Result<()>;

fireflow's rolling aggregation will keep groups in several generations and needs to check older ones without growing them.

• Implemented on all six interners: primitive, column, bytes, bytes_view, boolean, rows. Each mirrors its intern lookup without the insert path.
• ArrowBytesMap and ArrowBytesViewMap get a lookup-only find with the same null, hash and equality handling as insert_if_new.
• Required rather than defaulted. new_group_values picks the implementation, so a custom GroupValues can't be plugged into AggregateExec; the break only affects direct implementors.

Relates to apache#24603.

Tests

• A shared check per dispatched key type (Int32, Utf8, LargeUtf8, Binary, Utf8View, BinaryView, Boolean with null, (Int64, Utf8) → column, (List<Int32>, Int32) → rows): find matches intern, missing rows are None, and len() is unchanged.
• cargo test -p datafusion-physical-plan --lib group_values (84 passed), also with --features force_hash_collisions; clippy clean.