docs(tutorial): add executable GQA authoring tutorial

[zhen8838]

Why

• Add a concrete decode GQA authoring workflow without maintaining a second hand-synchronized program.

What

• Add six executable tutorial Modules and a Markdown renderer in attn_layer.py.
• Generate docs/tutorial/authoring.md from the source and register the page.
• Add the single exact-match test requested for source stdout and the committed page.

Contract

• Running ./tests/fixtures/tutorial/attn_layer.py emits the complete page on stdout.
• The generated page contains deterministic static analysis reports; no CUDA kernel is launched.

Risk

• The tutorial is static-analysis evidence only; it is not a runtime benchmark.