Handle backpressure

ESDB can be overloaded with read/write operations and it has no limitations enforced on the edge to prevent this.

The idea is to set the connection limit (in general or per operation) and deny requests when the limit is exceeded. It will improve the database stability, so issues with user's environment or apps would not take down the database.

Another aspects include handling back pressure with writes that come from elsewhere (projections, checkpointing, etc), not only from clients.

Dynamically react to growing replication window between cluster nodes could also be helpful to avoid slow recovery when one of the nodes is lost.