Huge Document demo and benchmark should share a query-param config contract

Problem

The Huge Document docs page and /dev/editor-perf were acting like related surfaces, but they did not share a config contract. That meant the manual demo could be set up one way while the benchmark harness quietly ran something else.

That is how perf work turns into folklore.

Solution

Create one shared Huge Document config module for the common knobs:

• blocks
• chunking
• chunk_size
• content_visibility

Use that shared module to:

• parse and write Huge Document query params
• parse and write benchmark-page query params
• build a direct Open in benchmark mode link from the docs page into /dev/editor-perf

Also make the benchmark page reuse the real mixed huge-document workload generator instead of maintaining a separate fake copy.

Why This Works

The docs page and the benchmark page now agree on the scenario before any measurement happens.

That gives you a clean workflow:

1. eyeball the scenario on the public docs page
2. jump straight into the benchmark surface with the same knobs
3. measure the real thing instead of recreating the setup by hand

Prevention

• If a docs example is also a benchmark reference, give it a query-param bridge into the actual harness.
• If two surfaces claim to represent the same workload, make them share the config parser and the workload generator.
• Never maintain two “almost the same” huge-document builders unless you enjoy debugging ghosts.