Separate retry policies for transient errors and setup errors

[istreeter]

If an error is due to bad warehouse setup (e.g. insufficient permissions) then we want a long slow retry delay, and we want the loader to send helpful error messages to the monitoring webhook.

If an error is due to a transient error (e.g. server error or network error) then we want a faster retry delay, and no need to send messages to the webhook. If the app crashes due to many successive transient errors, then the exception will get tracked in Sentry.