Data store clones can trigger CAF crash when master isn't locatable

[ckreibich]

When Zeek's manager isn't available and other nodes try to establish data store clones, it leads to Broker's "unable to locate master" error path. Under the right circumstances (for which I have no reproducer), this can trigger a CAF crash that looks like this:

caf::scheduled_actor::run_actions()+0x00000064
caf::scheduled_actor::quit(caf::error)+0x00000374
broker::internal::store_actor_state::send_later(caf::actor const&, std::chrono::duration<long, std::ratio<1l, 1000000000l> >, caf::message)+0x000002d4
broker::internal::core_actor_state::attach_master(...)+0x000018d8
non-virtual thunk to caf::detail::default_action_impl<caf::flow::op::from_resource_sub<caf::async::spsc_buffer<broker::cow_tuple<broker::topic, broker::internal_command> > >::run_later()::{lambda()#1}, false>::run()+0x000000e8
caf::scheduled_actor::run_actions()+0x00000184
caf::scheduled_actor::finalize()+0x00000028
caf::scheduled_actor::activate(caf::execution_unit*)+0x00000064
caf::scheduled_actor::resume(caf::execution_unit*, unsigned long)+0x00000054
std::thread::_State_impl<std::thread::_Invoker<std::tuple<caf::actor_system::launch_thread<caf::scheduler::worker<caf::policy::work_sharing>::start()::{lambda()#1}>(char const*, caf::scheduler::worker<caf::policy::work_sharing>::start()::{lambda()#1})::{lambda(auto:1)#1}, caf::intrusive_ptr<caf::ref_counted> > > >::_M_run()+0x000000fc
std::error_code::default_error_condition() const+0x0000003c
pthread_condattr_setpshared+0x000004f8
clone+0x0000005c