Add schema-mapped local data providers

[luisleo526]

Summary

• add runtime TabularSchema inspection and partial or explicit BarColumnMapping APIs
• add built-in CSV, read-only SQLite, and SQLAlchemy 2.x historical bar providers
• register csv, sqlite, and sqlalchemy for the existing raw-Pine backtest harness
• support numeric and ISO-8601 timestamps, optional symbol/timeframe columns, strict duplicate detection, and normalized provenance
• add environment-variable database URLs so credentials do not need to live in provider JSON
• document programmatic and CLI workflows for arbitrary file/table/view schemas

Why

Users should be able to backtest private datasets without migrating them into a PineForge-owned DDL. Runtime reflection plus explicit semantic mapping supports arbitrary source names while avoiding unsafe guesses and raw SQL configuration.

Safety and behavior

• common OHLCV aliases are inferred; missing or ambiguous mappings fail with available-column guidance
• SQLite opens files read-only
• reflected identifiers are validated and safely quoted; symbol, timeframe, and time bounds use bound expressions or parameters
• SQLAlchemy connection parameters are hidden from logging by default
• local records are treated as confirmed snapshots and still pass the normal PineForge Bar validation

Validation

• ruff format --check src tests
• ruff check .
• mypy src
• pytest — 82 passed, 2 expected Docker-gated skips
• python -m build
• clean-wheel import without SQLAlchemy installed, confirming the database dependency stays optional