Fix incorrect returning of stakes and regen SNL & SQL DB metadata

[Doy-lee]

Currently the network has the incorrect stake amounts to unlock to operators because the stake amount was not correctly multiplied by BATCH_REWARD_FACTOR when inserted into the SQL DB. The staking_requirement was also not serialized to disk (so once the node restarts that value gets set to 0) and when the exit is processed, the stake amount to return gets clamped to 0. Right now only 2 exits have been done on the network which are from my nodes.

This patch fixes that and increments the version of the SNL which will trigger everyone to recalculate the values in the DB and the metadata (and correctly store the staking_requirement).

If the network tries to claim now there'll be inconsistencies in the amount of rewards claimable by the operator (me) and the aggregation will fail. The more operators that upgrade the better as the BLS aggregation will have less non-signers and more nodes with this patch will agree on the correct amounts to return.

Additionally noticed another adjacent bug where if you had multiple exits in the one block, only the first stake would get queued for unlocking and the rest was dropped due to an insert conflict.

More specifically:

• The SQL DB was not deconflicting when inserting a delayed payment (e.g. returning stake after exit). It has primary key on (eth_address, payout_height) which meant that if you had multiple stakes returning on the same height, it'd just completely ignore the payment, and hence incorrectly return only the first stake that you unlocked. Fixed this by adding a deconflict step to sum the amounts if they do.

More broadly, at a higher level the primary key looks out of place since the table is not really relational but it did help catch this error and the end result of coalescing rows is more ideal than having individual rows for each payment.

• BlockchainSQLite::reset_database() was not dropping some tables causing left-over data to persist in the DB which caused the regeneration of data to fail. So we now drop the delayed_payments table and set the height to 0.

• Fix service_node_list::public_key_lookup not searching the recently_removed_nodes list. The code path that relies on this fix is purely cosmetic and is used to log to the log files, so no repercussions here.

• In the recently_removed_nodes instead of just pulling out individual fields from the service_node_info on erase, store the whole struct which is useful for populating the BLS endpoints that end-user applications consume and helps them maintain a consistent view of a node as it transitions out of the network.

• The exit code was not storing the stake amounts into the DB in the correct units. I've introduced a sql_db_money which strongly types the units so that the conversion of monetary units is explicit at the DB interface.

[jagerman]

The staking_requirement was also not serialized to disk (so once the node restarts that value gets set to 0) and when the exit is processed, the stake amount to return gets clamped to 0.

I'm not seeing this happen on current dev: staking_requirement seems to get preserved across a restart.

Edit: Oh, wait, I see: this is only for recently removed nodes (and I was looking at currently registered nodes).

[jagerman]

Resyncing mainnet failed the payment validation check; pushed a fix.

[Doy-lee]

Added the fixes and bumped the version to 11.0.5