Pruning cutoff should be calculated from LIB instead of from head

As we are currently pruning from the head, if the PRUNING_CUTOFF is a small number, or if there is a loss of liveness (LIB stops advancing), we run the risk of pruning data that would still be required after a reorg.

Applying PRUNING_CUTOFF from LIB solves the issue. 