refactor: declarative schema for configuration

[cmgzn]

Core Changes

• Add data_juicer/config/schema.py (938 lines): Define DJConfig Pydantic model as single source of truth for global config
• Refactor data_juicer/config/config.py: Migrate 673 lines of field definitions to schema module; retain only parser construction logic
• Enhance ConfigValidator: Support merging BASE_CONFIG_RULES with CONFIG_VALIDATION_RULES

Technical Benefits

• Centralized config field definitions, eliminating scattered parser.add_argument() calls
• Provide query APIs (get_json_schema(), get_defaults()) for external tool integration
• Type validation delegated to Pydantic; runtime checks moved to model definition phase

Impact Scope

• data_juicer/core/data/dataset_builder.py: Adapt to new config structure
• data_juicer/core/data/load_strategy.py: Adapt to merged validation rules
• data_juicer/tools/DJ_mcp_recipe_flow.py: Adapt to config API changes
• Test cases updated accordingly

[gemini-code-assist]

Code Review

This pull request introduces a declarative configuration schema using Pydantic, centralizing configuration definitions and refactoring the argument parser and default settings logic. It also updates dataset building and event logging to support the new configuration structure. Feedback highlights a redundant conditional block in the schema registration function and an unused helper function in the recipe flow tool that should be removed.