scm: connection peer resets under heavy load, incorrect event on error

scm: connection peer resets under heavy load, incorrect event on error

Under heavy load, we get a peer reset from the remote stack. In this
case retry the socket connection for this QP setup.

Add debugging with PID's and socket ports to help isolate
these types of socket scaling issues.

Report correct UD event during error, check remote_ah creation.

Fix dapl_poll return codes for single event type only.

Signed-off-by: Arlin Davis <arlin.r.davis@intel.com>