Transition when erroring due to duplicate work should be to client error

[willscott]

we had a case where a retrieval was picked up by a bot that was already stuck trying to do the same job.

ERROR: retrieval failed: Retrieve failed: there is an active retrieval deal with peer <peerid> for payload CID <payload> (retrieval deal ID <deal>, state DealStatusOngoing) - existing deal must be cancelled before starting a new retrieval deal

the logs in dealbot however were

2021-05-17T05:04:43.026Z        INFO    engine  engine/engine.go:117    successfully acquired task      {"uuid": "469616cc-9c04-426d-9869-4f1295a23a9c", "worker_id": 1}
2021-05-17T05:04:43.067Z        INFO    engine  tasks/retrieval_deal.go:66      got query offer {"root": "<payload>", "piece": "<nil>", "size": 34359738368, "minprice": "0", "unseal_price": "0"}
2021-05-17T05:04:43.073Z        INFO    engine  engine/engine.go:157    successfully retrieved data     {"uuid": "469616cc-9c04-426d-9869-4f1295a23a9c"}

We need that to be marked instead as a client error

[ghost]

Possible dup #156. We should talk about these two approaches sync.

[willscott]

we will no longer publish these failures to the public output of the dealbot, so they won't be picked up by reputation services. we don't have any explicit tagging of 'client' vs 'miner' error beyond that, it's for the downstream user of the logs to infer or decide.