Pending conflicts could be resolved by a wrong replicator

It might not be a common case that an application has multiple pull replicators (e.g. for different channels or filters) and each replicator has a different conflict resolver. Currently, when starting a replicator, the replicator will try to resolve any pending conflicts in the database regardless of whether the replicator pulled the conflict docs or not. As a result, a pending conflict could be resolved by a wrong conflict resolver of the replicator that hasn't pulled the doc.