refactor: Deduplicate scoring logic and cleanup unused code in scoring.py

[nice-bills]

While reviewing the validator scoring logic in talisman_ai/analyzer/scoring.py, I identified several opportunities to improve maintainability and code quality through refactoring and cleanup.

Changes:

1. Deduplicate Post Age Logic: Both compute_post_score and score_post_entry contain identical blocks of code for calculating post age and enforcing the 10-day cutoff rule. I propose extracting this into a helper function is_post_too_old(created_at).
2. Remove Dead Code: The function _build_canonical_from_dict appears to be unused within the repository and contains a docstring inconsistency (claims to take a dict but treats it as a PostClassification object).
3. Improve Type Safety: Add missing type hints for the analyzer arguments in score_post_entry and top_k_relevance_from_analyzer to ensure they are correctly identified as SubnetRelevanceAnalyzer instances.
4. Cleanup Component Calculation: Refactor score_post_entry to use the unified scoring components to ensure consistent behavior across different entry points.

I would like to submit a PR to address these items.