mt-backup

Modify signal flow for multithreaded backup so that the major 'control'
signals which control backup execution are sent to all LDM instances
instead of to LDM1 alone. These signals are DEFINE_BACKUP_REQ,
START_BACKUP_REQ, STOP_BACKUP_REQ and ABORT_BACKUP_ORD.

1) Add signal handling to BackupProxy to accept DEFINE_BACKUP_REQ,
START_BACKUP_REQ, STOP_BACKUP_REQ and ABORT_BACKUP_ORD. These signals
are forwarded to all LDMs.

2) Modify BACKUP to check backup type and send control signals
accordingly: send to LDM1 for single-threaded backup and to proxy
for multithreaded backup.

3) Add masterRef to DEFINE_BACKUP_REQ so that all participants set a
masterRef explicitly specified by the master.

4) Use senderRef instead of masterRef as destination for all
DEFINE/START/STOP_BACKUP_CONFs and REFs.

The masterRef is the BACKUP block which coordinates the backup
across all the nodes, i.e. LDM1 on the master node. The senderRef
is the BACKUP block which sent the last REQ signal. The masterRef
sends signals to the BackupProxies on all the nodes, and each
BackupProxy sends the signals to the LDMs. So the LDMs need to reply
to the BackupProxy, not the master.

- For single-threaded backup: backup master directly controls
participants on all nodes, so
  masterRef = senderRef = LDM1_on_master_node.
- For multithreaded backup: backup master sends control signals to
BackupProxy on each node + each BackupProxy controls backup exec
across LDMs, so:
  For all LDMs on node N, senderRef = BackupProxy_on_node_N
  For all LDMs on all nodes, masterRef = LDM1_on_master_node.

5) Add checks to detect and handle case where BackupProxy failed to
clean up SignalSenders from previous backups.