ceph_manager: take int or string to osd_admin_socket

This fixes a failure on dump_stuck.

teuthology/task/ceph_manager.py

@@ -270,7 +270,7 @@ class CephManager:
         remote = None
         for _remote, roles_for_host in self.ctx.cluster.remotes.iteritems():
             for id_ in teuthology.roles_of_type(roles_for_host, 'osd'):
-                if int(id_) == osdnum:
+                if int(id_) == int(osdnum):
                     remote = _remote
         assert remote is not None
         args=[