research: possible concurrency issue involving wait_for_commit variable

[aubrika]

During relayer debugging, we have seen a few instances of the nullifier XXX already spent in YYY error, indicating that the planner is trying to spend the same notes in different transactions. This seems to be an issue where build_penumbra_tx is called more than once, which has usually not been the case, as the observed volume has been able to process within a single transaction. This indicates that there may be an issue at scale and this should be ruled out - appropriate usage of wait_for_commit should be audited and confirmed.