Bucket System for Traffic Light Metric for Code Suggestions

Currently traffic light metric for Code Suggestions is measured using mean instead of buckets as pointed out here.

This issue is to switch to buckets instead and to follow the following logic:

Code Suggestion Using Cosine Similarity Scores

Scores	Color	Status
Above 80% of answers in range 0.8 and above	🟢	recommended
60-70% of answers in range 0.8 and above	🟡	recommended
below 60% of answers in range 0.8 and above	🔴	not recommended

Definition of Done

• Notebook is updated to use the bucket system instead
• Docs are updated with the new results