What deployment strategy should you implement to ensure Bigtable maintains read and write availability by automatically routing requests to the alternate cluster if either cluster-a in us-east or cluster-b in us-central1 becomes unavailable, given that data replicates from cluster-a to cluster-b?