feat(fuzzer): add multi-kernel fuzzing framework

[wangqin1723-max]

Add a standalone fuzzing framework for automatically generating and testing multi-kernel PyPTO programs. The framework generates random InCore kernel functions and combines them through Orchestration functions using different composition modes.

Components:

• Core fuzzer engine (fuzzer.py) for random test case generation
• Kernel generator for InCore functions with operator chaining
• Multi-kernel test generator with 3 composition modes:
  • Direct: Single kernel execution
  • Sequential: Kernel result chaining
  • Merge: Result aggregation
• Orchestrator generator for kernel coordination

Features:

• Support for 20+ operators (binary, unary, scalar, advanced)
• Configurable tensor shapes and operator weights
• Memory hierarchy support (L1/L0A/L0B/L0C)
• Automatic correctness verification with numpy reference
• Shape alignment validation for block operations

Documentation:

• README.md: Usage guide and examples
• OP_RULES.md: Operator rules and constraints
• CHANGELOG.md: Development history

Testing:

• Generated test cases (test_fuzz_multi_kernel.py)
• Example script (example_multi_kernel.py)
• Test case for expand operations (test_expand.py)