Add python_script node code generation with flowfile API rewriting

[Edwardvaneechoud]

Summary

This PR adds support for generating executable Python code from python_script nodes by rewriting flowfile API calls into plain Python equivalents. This enables python_script nodes to be converted to standalone functions that can run outside Docker kernel containers.

Key Changes

• New module python_script_rewriter.py: Implements AST-based analysis and transformation of flowfile API calls

  • analyze_flowfile_usage(): Parses user code to detect flowfile API patterns, artifact dependencies, and unsupported calls
  • rewrite_flowfile_calls(): Transforms flowfile.* calls to plain Python (e.g., flowfile.read_artifact("x") → _artifacts["x"])
  • build_function_code(): Wraps rewritten code into a function definition with proper parameters and return statements
  • extract_imports(): Extracts non-flowfile imports from user code
  • get_required_packages(): Cross-references user imports with kernel package requirements

• Updated code_generator.py:

  • Added _handle_python_script() method to process python_script nodes
  • Tracks published artifacts across nodes for validation of downstream dependencies
  • Validates artifact availability and detects unsupported API patterns
  • Emits kernel package requirements as comments
  • Added _published_artifacts and _has_python_script_nodes state tracking
  • Initializes shared _artifacts = {} store in generated code when needed

• Updated __init__.py: Exports new rewriter functions and classes for public API

Implementation Details

• Flowfile API Mapping:

  • flowfile.read_input() → function parameter input_df
  • flowfile.publish_output(expr) → return statement
  • flowfile.publish_artifact("name", obj) → _artifacts["name"] = obj
  • flowfile.read_artifact("name") → _artifacts["name"]
  • flowfile.log(msg, level) → print(f"[{level}] {msg}")

• Validation: Detects and reports:

  • Dynamic artifact names (non-string literals)
  • Unsupported API calls (display, publish_global, etc.)
  • Missing artifact dependencies from upstream nodes
  • Syntax errors in user code

• Code Generation: Produces properly typed function signatures with LazyFrame parameters and return types, maintaining compatibility with the polars-based pipeline architecture