rpc_client_call() forgets who it is calling if it's _client_connection[] is deleted by another thread

Create issue
Issue #167 resolved
dd1 created an issue

clients removed by bm_cleanup() & co seem to remain in the RPC connection cache and the next run transition gives this error message. Right before this, fectrl was hard-killed by stopping from the mhttpd "programs" page. Then I try to start a run from mhttpd.

21:54:23.440 2019/02/19 [mhttpd,ERROR] [midas.c:11761:rpc_client_call,ERROR] call to "" on "" RPC "rc_transition": timeout waiting for reply

21:54:23.439 2019/02/19 [mhttpd,ERROR] [system.c:4758:ss_recv_net_command,ERROR] timeout receiving network command header