feat: Add Scholarship & Mentorship scoring category and fix provider …

[vivekTiw120303]

Resolves #76

New Feature: Scholarship & Mentorship Scoring

This PR adds a new scoring category for scholarships and mentorships as requested in issue #76.

• Added Scholarship and ScholarshipsSection to models.py.
• Added the scholarship_mentorship score to the Scores model.
• Created a new scholarships.jinja prompt for data extraction.
• Updated pdf.py, transform.py, and template_manager.py to use the new prompt.
• Updated score.py to display the new category in the final report.

Bug Fixes Included

• Provider Selection: Fixed a bug in llm_utils.py and prompt.py where the LLM_PROVIDER environment variable was being ignored, ensuring the correct provider is now used.
• Dependencies: Updated requirements.txt to resolve a deep dependency conflict with Google's libraries that was preventing the Gemini provider from working correctly.

Testing Note

The feature has been fully implemented and tested with the default Ollama provider. From a code perspective, the feature is complete.

However, the local gemma3:4b model was not consistently powerful enough to pass the validation test (i.e., extract the scholarship data correctly). Furthermore, a persistent cloud configuration issue (404 Publisher Model not found) prevented successful validation with the Gemini/Vertex AI provider, even after a full environment overhaul and multiple debugging attempts. The included bug fixes for provider selection and dependencies were discovered during this process.