Now start_nodes works even with dropping repslots on ext cleanup.

But the problem of properly handling active repslots still requires attention.
We should probably add some func 'prepare_for_ext_cleanup' which will forcefully
destroy all such slots.

Author: arssher

bin/shardman_start.sh

@@ -7,7 +7,8 @@ cd "${script_dir}/.."
 make clean
 
 restart_nodes # make sure nodes run
-for port in $master_port "${worker_ports[@]}"; do
+# first workers, then master
+for port in "${worker_ports[@]}" $master_port; do
     psql -p $port -c "drop extension if exists pg_shardman;"
 done