Add RemoveOrphanFiles maintenance action

[manuzhang]

iceberg-cpp implements ExpireSnapshots (src/iceberg/update/expire_snapshots.{h,cc}) but does not have an equivalent of Java's RemoveOrphanFiles action. After expiration, table-scoped GC of unreferenced files (data files, delete files, manifests, manifest lists, statistics files, sidecar files) is left to the user.

Scope

• Walk the table location; collect every file path
• Walk all referenced files reachable from snapshots in TableMetadata.snapshots (data + delete files via manifests, manifest lists, statistics files, partition statistics files, metadata.json files referenced by metadata-log)
• Optional staleness threshold (skip files newer than N hours, mirroring the Java action's olderThan to avoid racing with concurrent writers)
• Pluggable delete_func similar to SnapshotUpdate::DeleteWith

Reference

• org.apache.iceberg.actions.RemoveOrphanFiles

[manuzhang]

Upstream status: partial — file-walk action still missing.

Java distinguishes two cleanup paths:

1. ExpireSnapshots cleanup — walks the metadata graph, deletes files unreferenced after expiration.
2. RemoveOrphanFiles — walks the filesystem, deletes files unreferenced by any active snapshot regardless of expiration.

Upstream has #1 only:

• apache/iceberg-cpp#592 (MERGED) — feat: add ReachableFileCleanup to expire snapshots
• apache/iceberg-cpp#648 (OPEN) — adds IncrementalFileCleanup strategy
• apache/iceberg-cpp#658 (OPEN) — bulk/parallel file deletion for ExpireSnapshots

No upstream issue or PR for the standalone table-walk RemoveOrphanFiles action. Recommendation: keep this open, file on apache reframed as "table-walk RemoveOrphanFiles action distinct from ExpireSnapshots cleanup."