Let update_elected_producers() reduce the schedule

[cc32d9]

Line 125 in https://github.com/EOSIO/eosio.contracts/blob/7c042002b903bbf13c7ae37851f03edf58ae9ff8/contracts/eosio.system/src/voting.cpp

      if( top_producers.size() == 0 || top_producers.size() < _gstate.last_producer_schedule_size ) {
         return;
}  

so, if a small network has less than 21 active producers, a producer cannot leave the schedule because the unregprod will be ignored. This is dangerous for network stability because a BP can unregprod and switch off the server, being sure that they did the right thing. The network will not be able to replace the BP or remove it from the schedule, unless they create sock puppets to fill in 22 positions.

I believe this rule should be removed. It's causing a danger for small networks and doesn't make sense for big ones.

It should be modified to disallow reducing the schedule to zero only.

[NatPDeveloper]

Running a small network, having this issue.

[DouglasHorn]

It would be better to make this number configurable and able to be modified by the existing BP vote as it presents a risk of network lockup in certain edge cases.

[deckb]

Moving to the new reference system repository.

There is a more general configuration for the system contract that should be implemented for this contract. I believe we can roll this into that change.