docs: final fork-budget numbers after #817

[Chemaclass]

🤔 Background

The fork-budget rule's numbers predate #817 (provider map single build); stale budgets would mislead the next perf pass.

💡 Changes

• Sequential budget corrected to 2 awk; session outcome recorded (acceptance suite ~61s → ~17s wall, CPU halved across perf(runner): drop 6 mktemp forks per cold start (#798) #801-perf(runner): build the provider map once per file, not twice #817).
• Measured --coverage note: remaining forks are bounded by unique-file count (nightly, non-gating) — documented as not worth chasing.