OpenBGPD 9.0

• Rewrite the Adj-RIB-Out handling to be more memory efficent and faster. For large IXP route server deployments a reduction in memory usage of more than 50% should be feasible.
• Process UPDATE messages in two phases: first update Adj-RIB-In, Loc-RIB, and FIB, then process all the Adj-RIB-Out tables. This significantly reduces the latency since updating all the Adj-RIB-Out tables could take a fair amount of time.
• Introduce CH hash tables - a scalable hash map implementation that boosts performance through improved cache locality.
• Introduce new metrics that track the amount of time spent in various parts of the main event loop of the route decision engine.
• Fix various non-criticial things uncovered by Coverity scanner.