Re-fetch on a 409 when two Oban jobs create the same record

When two Oban jobs GET the same record at once, both get a 404 and both POST it. The second POST gets a 409 Conflict. That 409 means the record exists now, so fetch it again and carry on.

We returned the 409 as an error instead. Oban retried the job until a later GET found the record. The job got there in the end, but each retry reported an error to AppSignal.

The fix matches the conflict right after the create. Request.create/1 is the POST, and request_get/1 is the same GET the job started with:

params
|> Request.create()
|> case do
  {:error, [%{"title" => "Conflict"} | _]} ->
    # another job created it between our GET and POST
    request_get(params)

  result ->
    handle_result(result)
end

That match takes any 409 Conflict as “already exists”. On an endpoint that can conflict for another reason, match the error detail too.