provide bfd based liveliness detection for vmis

Currently, one can enable health-check based on ping and curl commands. When enabled, these are run periodically, once every few seconds. Hence failure detection times using these can be quite large, always in the order of seconds.

Use BFD to detect failure and recovery instead of ping/curl. BFD runs quite aggressively and typically can detect end point failures in the order of milli seconds. When used in an event (notify) driven model (instead of poll), one can achieve failure detection and recovery in sub-second intervals