OSD: avoid failure peer info to resent

OSD: avoid failure peer info to resent

  maybe_update_heartbeat_peers may remove one peer and never add in.
  if that peer is in failure_pending but that peer is not really has problem,
  failure_pending will hold that peer until it really down or add in again.

  once ms_handle_connect called, these pending failure will be resent again.

  2019-09-12 09:44:47.080933 7f1fad781700 10 osd.13 6175 ms_handle_connect on mon
  2019-09-12 09:44:47.080937 7f1fad781700 10 osd.13 6175 send_alive up_thru currently 6159 want 6155
  2019-09-12 09:44:47.080945 7f1fad781700 10 osd.13 6175 requeue_failures 0 + 1 -> 1

Signed-off-by: 宋顺10180185 <song.shun3@zte.com.cn>