Fix quality_avg to compute mean final_score instead of completion rate

[Copilot]

quality_avg was computing completion rate (completed_count / total * 100) rather than the average of actual quality scores, making it a misleading metric.

Change

• Extract final_score values from completed traces where final_score is not None
• Compute arithmetic mean of those scores as quality_avg

# Before — completion rate, not quality
sum(1 for t in recent_traces if t.status == "completed") / max(len(recent_traces), 1) * 100

# After — true average quality score
completed_quality_scores = [
    t.final_score
    for t in recent_traces
    if t.status == "completed" and t.final_score is not None
]
quality_avg_value = sum(completed_quality_scores) / max(len(completed_quality_scores), 1)

final_score is already on a 0–100 scale, so no additional scaling is applied. Falls back to 0.0 when no scored traces exist.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.