UpdateSchema: moving a nested column added in the same transaction is silently ignored

### Apache Iceberg version

main (development)

### Please describe the bug 🐞

When a nested column is added via `AddColumn` / `UnionByNameWith` and then reordered with `MoveFirst`/`MoveBefore`/`MoveAfter` in the same `UpdateSchema`, the move is silently discarded (or fails to resolve a sibling), and the applied schema keeps the field at its default appended position.

Reason for this: `parentID` map, which is seeded once in `initParentID` is never updated. `moveColumn` to `findParentID` therefore returns `TableRootID (-1)`, thus filing under the root struct and dropped by `moveFields` during apply.

Top-level added columns are unaffected.

### Contribution

Will submit a PR for the fix.

Start by tracing initParentID, moveColumn, findParentID, and moveFields in the UpdateSchema implementation. Reproduce an UpdateSchema that adds a nested column and then calls MoveFirst, MoveBefore, or MoveAfter; done means the move resolves the nested parent and the applied schema preserves the requested position.

Written by the indexing model from the issue text.