-Go back to restarting the log in process if rakp2 or rakp4 goes awry, it's too likely for a dropped packet to confuse IPMI 2 devices.
-Fix relog to reset session context to avoid getting distracted by now stale RAKP messages if the IPMI device happened to be slow
-Switch session ids on relog, so that everything can ignore packets from a zombie IPMI session reliably (e.g. relog recovery of a session).
-In parse_ipmi_payload, delete pendingargs in addition to deleting sessions_waiting entry. The latter should have sufficed, but there persisted an odd
issue where sometimes the last non-logout packet was erroneously retransmitted as retries at the end. This change short-circuits
the timedout function should this occur, generally delaying exit by 100 milliseconds whet it goes wrong.
git-svn-id: https://svn.code.sf.net/p/xcat/code/xcat-core/branches/2.7@13821 8638fb3e-16cb-4fca-ae20-7b5d299a9bcd
ba244e8841