[BP 6.6.2] Excessive xdcr connections (could be stale connections)

While analyzing the CBSE-7988, the following issues are observed.

1) Unclean Pipeline start timeout and restart: In certain situation the pipeline start process can cause tcp connection and go routines leak. The pipeline start spawns up routines and tries to create xmem nozzle and other parts of the pipeline in parallel. If operation like xmem nozzle create takes sometime, the pipeline start timeout can leave the connections and routines in dormant and move on without closing them.   

2) excessive tcp connections: #1 can cause connection leak.

3) Unaligned timeout dependencies. The pipeline start executes the different operations like xmem nozzle creation, DCP nozzle creation... each operation has its own timeout profile. i.e. the pipeline start process has 30sec timeout however the sub operations like xmem nozzle creation can take 120sec. This unaligned timeout dependencies can lead to inconsistency