Local shovel: handle unacked messages in queue down

[dcorbacho]

It affects messages routed through exchanges, when there is no target queue configured in the shovel (it could be many queues as destination or none).

[michaelklishin]

@dcorbacho the CI failures seem repeatable, even though I cannot reproduce them locally:

local_dynamic_cluster_SUITE > tests > local_to_local_dest_down
    #1. {error,{amqp10_event_timeout,link,
                                     {link_ref,sender,<0.2678.0>,0},
                                     attached}}

local_dynamic_cluster_SUITE > tests > local_to_local_multiple_dest_down
    #1. {error,
            {{shutdown,
                 {connection_closing,
                     {server_initiated_close,541,
                         <<"INTERNAL_ERROR - Could not declare queue 'local_to_local_multiple_dest_down_src' in vhost '/' on node 'rmq-ct-local_dynamic_cluster_SUITE-1-21000@localhost' because the metadata store operation timed out">>}}},
             {gen_server,call,
                 [<0.2695.0>,
                  {call,
                      {'queue.declare',0,
                          <<"local_to_local_multiple_dest_down_src">>,false,
                          true,false,false,false,[]},
                      none,<0.2682.0>},
                  infinity]}}}

local_dynamic_cluster_SUITE > tests > local_to_local_no_destination
    #1. {error,
            {{shutdown,
                 {connection_closing,
                     {server_initiated_close,541,
                         <<"INTERNAL_ERROR - Could not declare queue 'local_to_local_no_destination_src' in vhost '/' on node 'rmq-ct-local_dynamic_cluster_SUITE-1-21000@localhost' because the metadata store operation timed out">>}}},
             {gen_server,call,
                 [<0.2717.0>,
                  {call,
                      {'queue.declare',0,
                          <<"local_to_local_no_destination_src">>,false,true,
                          false,false,false,[]},
                      none,<0.2704.0>},
                  infinity]}}}

[dcorbacho]

@michaelklishin Fixed, the cluster needs 3 nodes when using Khepri. Changing the metadata store can be reproduced locally.