feat: Added role fit scoring

[PranuPranjal]

✨ Add Optional Role Fit Scoring to Hiring Pipeline

📝 Overview

This update introduces an optional Role Fit scoring system (0–20) that evaluates how closely a candidate’s profile aligns with a given role description. When a role file is provided, the system computes the score, displays evidence in the console, and exports results to CSV.

---

🎯 Motivation

Recruiters often need to assess role-specific suitability beyond generic resume quality.
This feature enables transparent, evidence-backed comparisons between candidates for a particular job description.

---

🔧 Changes Included

• models.py → Added role_fit: CategoryScore to Scores.
• prompts/templates/role_fit.jinja → New template defining a 0–20 rubric and JSON schema.
• prompts/template_manager.py → Loads role_fit.jinja.
• evaluator.py
  • ResumeEvaluator now accepts a role_description.
  • Performs structured LLM evaluation for Role Fit and attaches result to scores.role_fit.
• score.py
  • CLI now supports an optional role file:

    python score.py <resume.pdf> <role.txt>

  • Prints “Role Fit” category in console and ignores None categories in aggregation.
• transform.py → Exports role_fit_score and role_fit_max to resume_evaluations.csv.
• README.md → Updated with Role Fit usage instructions.

---

⚙️ How It Works

1. If a role description is provided, evaluator.py renders role_fit.jinja with the resume + GitHub/blog context and the role description.
2. The model returns structured JSON:

   {
     "score": <int>,
     "max": 20,
     "evidence": "<string>"
   }

3. The result is appended to evaluation.scores.role_fit.

Example CLI usage:

# Existing
python score.py /path/to/resume.pdf
With Role Fit
python score.py /path/to/resume.pdf /path/to/role.txt

Output:

• Console: Adds a “Role Fit” category with score and evidence.
• CSV: Includes role_fit_score and role_fit_max columns.

---

🔄 Backward Compatibility

• Fully backward compatible.
• If no role description is provided, Role Fit evaluation is skipped and existing behavior remains unchanged.

---

🧪 Testing

• Verified end-to-end with and without role file.
• No regressions in existing scoring categories.
• Role Fit appears only when applicable.
• CSV correctly includes new columns.

---

Closes #158