in-kernel PM: check to create new subflows when some are destroyed

[matttbe]

(linked to #496)

When a subflow is destroyed, it might be interesting to create a new subflow, e.g. we only tried to create one additional subflow because of the limits, but it was not possible to establish it for some reason (e.g. the MPJ+SYN got reset), it might be good to try again if it is possible to create more subflows.

It is not the first time someone reports that subflows are not created even if the limits are, according to the number of established subflows -- not what they can see with ss -M --, not reached.

Note: we should make sure not to loop: subflow between addresses (a) and (b) could not be established for some reason, we should not try to recreate it. → storing a bitmap of incompatible IDs?