refactor(udaf): name synthesized state fields state_{i}

The Python constructor path collapses input `Vec<DataType>` into
synthesized `FieldRef`s; bare integer names (`"0"`, `"1"`) made EXPLAIN
and debug output for the state schema opaque. Use `state_{i}` to match
the `arg_{i}` convention the codec applies on the input-field side.

The post-decode path is unaffected — the codec ships full
`FieldRef`s (with the names the sender chose) through IPC and bypasses
this synthesis entirely.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: timsaucer

crates/core/src/udaf.rs

@@ -180,7 +180,7 @@ impl PythonFunctionAggregateUDF {
         let state_fields = state_types
             .into_iter()
             .enumerate()
-            .map(|(i, t)| Arc::new(Field::new(format!("{i}"), t, true)))
+            .map(|(i, t)| Arc::new(Field::new(format!("state_{i}"), t, true)))
             .collect();
         Self {
             name,