Fix LogIndex Prune

[somnathb1]

Earlier assumption was that PruneLogIndex would only have to deal with block numbers after the last prune. So it could work with tx.cursor.First(). Now, after prune hack, kv.TransactionLog contains older logs for noPruneContracts, so, it must seek to the point of last prune progress.

Without this, for every update of head block, the stages would run incrementally, but LogIndex Prune would run from the beginning of TransactionLog table, spending 1-2 hours.