poteto-mode: give the perf playbook its eight strategy families

[poteto]

Summary

• The public perf playbook never received the strategy-families list that the private version's step 2 carries — its step 2 was a single line. This ports the full list at current parity: the six original families (divide-and-conquer, caching, indirection, batching, redundancy, lazy) plus the two just added upstream after checking the playbook against Hanson/Crain's performance mantras (anysphere/everysphere#168492):
  • Elimination ("don't do it") — leads the list; the one family a trace can't surface, so it rides the how pass.
  • Scheduling ("do it when they're not looking") — after Lazy, with the Lazy-vs-Scheduling distinction spelled out; measured on perceived latency.
• Text is verbatim from the private playbook minus one internal skill-path reference (already absent in the public step structure). Leak check clean.
• Deliberately NOT touched: the playbook's gpt-5.5-high-fast default — it's pstack's current configured default across 13 files, and model routing changes are their own PRs.

Test plan

• Leak check clean
• Step numbering and reply contract unchanged

---

Note

Low Risk
Documentation-only playbook text; no runtime, routing, or measurement pipeline changes.

Overview
Expands step 2 of the public perf-issue playbook from a single how/trace line into the full eight strategy families list used upstream, framed as trace-driven hypothesis generators (not a checklist).

Adds Elimination (work that may not need to run; needs the how pass, not only the profiler) and Scheduling (move work off the interactive moment, with Lazy vs Scheduling and perceived-latency measurement called out), alongside the six existing families: divide-and-conquer, caching, indirection, batching, redundancy, and lazy evaluation. Step numbering, measurement workflow, and the reply contract are unchanged.

^{Reviewed by Cursor Bugbot for commit b0785e1. Bugbot is set up for automated code reviews on this repo. Configure here.}