Use a dedicated CNode::m_avalanche_message_fault_score member variable to track repeated avaresponse message faults, and call Misbehaving(*peer, 100) when that score reaches a threshold instead of incrementing the misbehavior score by 2 directly.
In a case in which the only source of misbehavior is the avaresponse faults, the behavior is unchanged. If the misbehavior score is incremented for other reasons, now we are a bit more tolerant of network faults.
This prepares the codebase for a following diff (core#29575) in which we will stop tracking the misbehavior score, and calling Misbehaving will immediately trigger discouragement.
Depends on D18021