raft: configurable rejection of raft leadership

[dotnwat]

A core primitive in rolling upgrades is graceful shutdown of a node such that availability of the system in unaffected when the node goes offline. Since all cluster interaction occurs on nodes with leadership, expunging leadership from a node will result in clients naturally being redirected to other nodes in the system, and avoid disruption like raft time and leadership election that occur in non-graceful shutdown scenarios.

1. Applies to all raft groups (controller, internal topics, Kafka topics)
2. Leadership should be rejected for normal raft mechanics and explicit requests (e.g. from the balancer)
3. This mode should be able to be enabled and disabled

Additional notes

• RFC https://docs.google.com/document/d/1FcHekvM62RYRZqZaIvHbKucIQSGuSU2uuP6RW5r-oko/edit#
• PRD https://docs.google.com/document/d/140T_4Rg9cfIaTuPNlsEW-nfZ3INiLBHUxFy4-HsB3CY/edit#

[dotnwat]

Duplicate of #3016