Replicator.resetCheckpoint does not work as documented

The method `Replicator.resetCheckpoint` claims that it can be used on a Replicator that is stopped.

In fact it can only be used before the native companion object is (lazily) created (that happens, for instance, on a call to `getDocumentIds`)

This used to be the same thing as being Stopped. Now that a replicator can be restarted, though, it is not.

The following will cause an exception:

• create a replicator
• run it
• wait for it to stop
• call Replicator.resetCheckpoint

as will

• create a replicator
• ask for pending document ids
• call Replicator.resetCheckpoint

The error message will be the standard ReplicatorNotStopped message, even though the replicator is stopped.

Fixing this will require plumbing a new call to reset the replicator options, through the JNI