This PR implements a comprehensive refactoring to centralize PythonCall.jl's public API into a dedicated module structure.

Key Changes

New Centralized API Module

• Created src/API/ directory with organized API components:
  • types.jl - All public type definitions
  • functions.jl - Public function declarations
  • macros.jl - Public macro declarations
  • exports.jl - Centralized export statements
  • publics.jl - Public API declarations using modern Meta.parse() pattern

Type System Reorganization

All major types moved to centralized API:

• Core types: Py, PyException, pybuiltins
• Wrapper types: PyArray, PyDict, PyIO, PyIterable, PyList, PyPandasDataFrame, PySet, PyTable
• JlWrap types: PyObjectArray, PyObjectVector, PyObjectMatrix (with updated Ptr{Cvoid} storage)
• Convert types: Priority enums for conversion system

Function and Macro Centralization

• JlWrap functions (pyfunc, pyclassmethod, pystaticmethod, pyproperty, etc.)
• The pytable function from Compat
• The @py macro from PyMacro
• The @pyconst macro from Core
• All Core builtin functions

Module Structure Updates

• Individual modules now import from centralized API rather than defining their own types
• Eliminated local exports in favor of single-source exports
• Consistent import patterns across entire codebase

Technical Improvements

• PyObjectArray uses Ptr{Cvoid} instead of C.PyPtr for improved type safety
• Clean public API declarations using modern Julia patterns
• Removed redundant exports and simplified module dependencies

Impact

This refactoring establishes a clear architectural boundary between PythonCall's public API and internal implementation. It reduces code duplication, improves maintainability, and provides users with a single, well-defined interface while preserving full backward compatibility.

The centralized approach makes the codebase more modular and easier to extend, setting up a foundation for future API evolution.