docs(qdp): improve generated Markdown rendering

[viiccwen]

Related Issues

Part of #1358
Related to #1361

Changes

• Bug fix
• New feature
• Refactoring
• Documentation
• Test
• CI/CD pipeline
• Other

Why

QDP source docstrings contain reStructuredText roles and unfenced usage examples that do not render cleanly in the generated Docusaurus API reference.

This causes references such as :meth: and :class: to appear literally and makes Python usage examples harder to distinguish from surrounding prose.

How

• Replace reStructuredText roles with Markdown-compatible inline code in QDP docstrings.
• Wrap the QdpBenchmark usage example in a fenced Python code block.
• Wrap the QuantumDataLoader usage example in a fenced Python code block.
• Keep the changes documentation-only without modifying runtime behavior or the public API surface.

The changes are limited to:

• qdp/qdp-python/qumat_qdp/_backend.py
• qdp/qdp-python/qumat_qdp/api.py
• qdp/qdp-python/qumat_qdp/loader.py
• qdp/qdp-python/qumat_qdp/tensor.py

Validation

• Generated the QDP Python API reference with pydoc-markdown.
• Confirmed the usage examples render as Python code blocks.
• Confirmed reStructuredText role markers no longer appear in the generated Markdown.
• Existing QDP Python tests: 7 passed, 6 skipped because the native extension was not built.

Checklist

• Added or updated unit tests for all changes — not applicable; documentation-only change
• Added or updated documentation for all changes

[rich7420]

LGTM. Diff is docstring/Markdown formatting only, no logic changes. ruff check/ruff format --check pass. pydoc-markdown output has balanced fences and zero remaining RST roles. Related tests (testing/qdp_python/, testing/qumat/test_qdp_module.py) pass, 113 passed.

Minor: loader.py/api.py usage examples call torch.from_dlpack without importing torch. Pre-existing, not introduced here. Worth a follow-up fix, not blocking.