Evaluation: Add verdict badges for scores

[AkhileshNegi]

Is your feature request related to a problem?
Each numeric judge-metric score has a traffic-light verdict band, but the UI only displays the raw number. This requires users to remember cutoffs to interpret scores correctly.

Describe the solution you'd like

• Implement a verdict badge component with three bands: Good (4–5), Needs Refinement (2–3), Needs Improvement (0–1)
• Display it alongside each metric score and on the overall run-level verdict
• Exclude badges for cosine scores and unscoreable / N/A entries

Original issue

Is your feature request related to a problem?
Each numeric judge-metric score carries a traffic-light verdict band, but the UI shows only the raw number — so a reader has to remember the cutoffs to know whether a 3 is fine.

Describe the solution you'd like

• Add a verdict badge component with the three bands: Good (4–5), Needs Refinement (2–3), Needs Improvement (0–1)
• Render it next to each metric score and on the run-level overall verdict
• Render no badge for cosine scores and unscoreable / N/A entries, which carry none

Parent: #265