scylla-5.0.0+criteo2: gossip: Speed up wait for gossip settle

In a large cluster, a node would receive frequent and periodic gossip
application state updates like CACHE_HITRATES or VIEW_BACKLOG from peer
nodes. Those states are not critical. They should not be counted for the
_msg_processing counter which is used to decide if gossip is settled.

This patch fixes the long settle on every restart issue reported by
users.

Refs #10337

Closes #10892

scylla-5.0.0+criteo1

Fix incompatibility between 4.6 and 5.0

scylla-4.6.3+criteo5

Fix node failing to join big clusters

Full Changelog: https://github.com/criteo-forks/scylla/commits/scylla-4.6.3+criteo5

scylla-4.6.3+criteo4

Ensure gossip is enabled on all shards before starting the failure_de…

scylla-4.6.3+criteo3

Ensure gossip is enabled on all shards before starting the failure_de…

scylla-4.6.3+criteo2

Add attributes support

Code is mostly taken from the standard role manager
This was done to be on par with the standard role manager implementation

scylla-4.6.3+criteo1: deprecate internal_distributed_timeout_config

It has been replaced by internal_distributed_query_state

scylla-4.4.6-wantimeout1

Add observability on the throttling and background

scylla-4.5.3+criteo2

deprecate internal_distributed_timeout_config

It has been replaced by internal_distributed_query_state

scylla-4.5.1+criteo2

deprecate internal_distributed_timeout_config

It has been replaced by internal_distributed_query_state