Audit remaining notebooks for executable CI coverage

#2163 adds an indexing notebook and brings the regular test-docs execution list to six notebooks: import smoke test, temporal predicates, hop bounds, indexing examples, validation fundamentals, and static rendering. All six executed successfully in CI. Read the Docs renders notebooks with nbsphinx_execute = "never"; its build alone does not validate execution.

Audit the remaining tracked notebooks for useful execution coverage. Executing the validation tutorial exposed an obsolete API argument and missing-column examples that needed explicit strict mode, so this audit should check the promised behavior as well as successful execution.

Initial candidates and constraints:

• demos/gfql/benchmark_filter_pagerank_cpu_gpu.ipynb: its presentation helper reads ignored local benchmark receipts under plans/. Make its required data available in a clean checkout, or provide a small self-contained CPU example. Preserve verified benchmark provenance when displaying measured numbers.
• demos/demos_databases_apis/graphviz/plot_static_demo.ipynb: overlaps the covered rendering tutorial and can fall back to printing code when pygraphviz is absent. Any CI coverage should require real rendering.
• demos/more_examples/graphistry_features/layout_tree.ipynb: large random graphs and very large figures need a bounded, deterministic CI mode.
• GPU memory, CPU/GPU benchmark, and remote GFQL/Python notebooks need suitable hardware, datasets, or authenticated services. Identify dedicated lanes and explicit reasons for notebooks that remain outside regular CPU docs CI.

Acceptance criteria:

• Inventory notebook paths, dependencies, external data/services, expected runtime, and existing coverage.
• Execute promising CPU candidates from a clean checkout using the docs environment; repair stale examples and add meaningful result/error assertions where appropriate.
• Add verified candidates to the existing execution flow without making docs CI depend on large downloads or unavailable hardware/services.
• Record coverage or an explicit exclusion reason for the remainder, with separate follow-ups for GPU/service lanes and fixture cleanup.
• Confirm CI logs show actual execution and fail on notebook errors; rendering or optional-dependency fallback alone does not count.