graceful failover deletes bucket with force=true and can lead to poor delta recovery

The RCA for MB-35326 showed that a dead->replica state change occurred soon before the bucket was deleted, the replica state change was never flushed because a force delete was used.

* see comment here

The result of this is that KV at the restart/bucket creation, warmup will ignore dead vbuckets and as such, the vbuckets which could of been warmed up (because they were correctly flushed as active or replica) must now be rebuilt.

Raising this MB as it seems a force bucket delete seems to be counter productive for a graceful failover->delta recovery workflow.

Note-1: force shutdown noticed in ns_server.debug.log

[ns_server:info,2019-07-27T23:34:56.175-07:00,ns_1@172.23.105.47:ns_memcached-default<0.20811.9>:ns_memcached:delete_bucket:770]Deleting bucket "default" from memcached (force = true)

 
Note-2: Assuming graceful failover as it's mentioned in logging e.g.

[ns_server:debug,2019-07-27T23:34:56.831-07:00,ns_1@172.23.105.47:ns_config_log<0.199.0>:ns_config_log:log_common:231]config change:

counters ->

[{'_vclock',[{<<"66664331ef97ea7bd2fdcdbe29e89002">>,{54,63731514896}}]},

 {failover_complete,2},

 {graceful_failover_start,2},

 {rebalance_success,24},

 {rebalance_start,24},

 {graceful_failover_success,1},

 {failover,1}]