WIP Changes in patterns related to experimental Hivemind that for paidout...

WIP Changes in patterns related to experimental Hivemind that for paidout posts takes rshares data from compatible hived rather than compute it itself.
Up to some HF hived allowed votes to be recast as long as they were in different cashout windows - previous HM treated them as edits.
Votes cast before HF1 were effectively scaled by 1mln - previous HM didn't take that into account.
Fixed bug with dust votes treated as downvotes, but it won't show before HF20.
Accepted differences due to above:
 - total_votes - new values bigger
 - net_votes - new values both ways
 - abs_rshares - new values bigger
 - net_rshares - new values both ways
 - flag_weight - since negative votes cast before HF1 were now corrected into more negative - new values bigger
 - rshares in votes for votes cast before HF1
 - vote value estimation and sometimes also score in notifications
Due to fixed bug with votes incorrectly attributed to recreated posts instead of their deleted version, id of notifications as well as number of unread were affected.
Since now HM receives vop for all payouts, not just final or nonzero, it was possible to fix last_payout value for partial payouts with no rewards.
Note: in order to reduce number of concurrent changes vote scaling was not included in reputation calculations, but formally it should (maybe later)