remove indexation_state::end method

there is no need to dump all cached blocks (not yet dumped to the database ) when
the plugin is closing. Because HAF has never more blocks dumped than
node has processed, then  we don't need to flushing all cached blocks when
the node is going to close, they will be dumped with replay after the
node will start.
In case of reindex some part of irreversible
blocks may stay inconistent ( they are not claimed by the randevouz
trigger), they won't be achived by the HAF applications and after restart
of the node they will be removed, and the rest of blocks from the node state
will be dumped during continuing the reindex.. Similar situation occurs with P2P
sync.
In case of LIVE sync the situation is much simpler: each block must be
flushed to stay in the node state, and if a block stays in the cache it
means it is not already confirmed in the state.