Fix compact dated LiteLLM model refs

[Kcstring]

Fixes #333

Summary

• normalize trailing compact date model refs like gpt-5.2-20251211 as a fallback after exact model lookup fails
• keep the original cache key so repeated lookups of the original ref still benefit from caching
• cover both find_provider_model and the public calc_price path for LiteLLM-prefixed refs

Tests

• UV_CACHE_DIR=/tmp/uv-cache UV_PYTHON_INSTALL_DIR=/tmp/uv-python uv run --package genai-prices pytest tests/test_model_matching.py -k litellm_provider_id
• UV_CACHE_DIR=/tmp/uv-cache UV_PYTHON_INSTALL_DIR=/tmp/uv-python uv run --package genai-prices pytest tests/test_price_calc.py -k compact_date_model_ref
• UV_CACHE_DIR=/tmp/uv-cache UV_PYTHON_INSTALL_DIR=/tmp/uv-python uv run --package genai-prices ruff check packages/python/genai_prices/data_snapshot.py tests/test_model_matching.py tests/test_price_calc.py
• UV_CACHE_DIR=/tmp/uv-cache UV_PYTHON_INSTALL_DIR=/tmp/uv-python uv run --package genai-prices ruff format --check packages/python/genai_prices/data_snapshot.py tests/test_model_matching.py tests/test_price_calc.py
• UV_CACHE_DIR=/tmp/uv-cache UV_PYTHON_INSTALL_DIR=/tmp/uv-python uv run --package genai-prices basedpyright packages/python/genai_prices/data_snapshot.py tests/test_model_matching.py tests/test_price_calc.py

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 3 additional findings.

[alexmojaki]

what about putting this logic in provider.find_model instead?