Different modes when using multiple upstream resolvers

[qdm12]

• "split": split at random every request between each upstream resolvers (current mode). BUT detect when a resolver fails consecutively compared to other resolvers, and turn it off for a certain period before trying again, so this prevents i.e. cloudflare from failing 50% of your queries if you have 2 providers set.
• "fallback": fallback to next resolver if the current one fails. Retry after some time with a dummy request to check if the ones "before" the current fallback resolver are back to normal, and go up the chain until we're back to the first resolver of choice. This is for latency worried users, where they really want 100% of their queries to go to provider number 1, but fallback to 2 if 1 doesn't work anymore.