Lease failure when update takes longer than socket timeout

The DocumentNodeStore may consider a lease update as failed when the operation takes longer than the socket timeout and later succeeds on MongoDB within the lease time. In this situation the lease mechanism currently thinks there was a lease update from another process and therefore backs off.

The implementation keeps track of the last successful lease update and based on this information performs a conditional lease update. The implementation should be more tolerant when it encounters an 'unexpected' update within the lease period and still consider the lease valid.