compiler: add Relu operator support

[strimo378]

Motivation

• Support the ONNX Relu operator end-to-end so small models with ReLU can be compiled and emitted as C.
• Provide both runtime evaluation and emitted C code that match ONNX Runtime behavior.
• Keep unary operator lowering consistent with existing Tanh handling and avoid adding extra C runtime deps.

Description

• Map the Relu op to a unary symbol by adding Relu → "relu" in _unary_op_symbol.
• Implement runtime evaluation for ReLU by adding np.maximum(value, 0) in _apply_unary_op.
• Update the Jinja2 unary operator template templates/unary_op.c.j2 to emit an inline ternary clamp expression for relu instead of a function call.
• Add tests and a golden C output: extend tests/test_endtoend_ops.py and tests/test_golden.py with ReLU cases and add tests/golden/relu_model.c as the expected codegen output.

Testing

• Ran the full test suite with pytest -q and all tests completed successfully.
• Test run result: 18 passed (tests include golden comparisons and compiled C testbench execution).

---

Codex Task