At 500 anonymous code pages a second the forge refuses none and answers p95 under 100 ms, with and without the edge cache #9514

From #9405 box 4.

Done looks like:

• [ ] bench/load.ts at 500/s, against a copy holding production-shaped data, reports zero 429s or 5xx and p95 under 100 ms. It is run once through the edge cache and once with the cache bypassed, and both results are kept in bench/results.
• [x] Whatever change gets there is named in the pull with its before and after numbers.
• [x] #9504's per-release compare carries this rate, so a later drop of more than 20% fails the release.

Files: src/http/ code-page routes, bench/results/, docs on load in docs/features/. Risks: the measurement is taken from a machine that is not the owner's Mac. A run that only reaches best case on a small copy is reported as that, not as done.

Reviewed pull #9548 at its head: ticked Whatever change gets there is named in the pull with its before and after numbers.; #9504's per-release compare carries this rate, so a later drop of more than 20% fails the release.. left open: bench/load.ts at 500/s, against a copy holding production-shaped data, reports zero 429s or 5xx and p95 under 100 ms. It is run once through the edge cache and once with the cache bypassed, and both results are kept in bench/results. — Only an in-process bypass run is kept; the run through the edge on a copy with production data is not taken (#9565)..

Left open after pull #9548 landed: 1 unticked item remains (bench/load.ts at 500/s, against a copy holding production-shaped data, reports zero 429s or 5xx and p95 under 100 ms. It is run once through the edge cache and once with the cache bypassed, and both results are kept in bench/results.). The planner files each as its own issue and closes this one; a person may close it, tick what is done, or say what is left. Until then the crew does not pick it again.

Split: the open box is now #9622, which waits on #9566 so it can be measured from a warm forge.