docs(AGENTS.md): when the frame budget outweighs a shared predicate

[derek73]

Adds one bullet to AGENTS.md's 2.0 conventions: frames are a budget for hot and scaling paths; everywhere else, prefer the shared predicate.

Why. The conventions asked for frame cost on hot paths but gave no way to weigh a frame against readability. On #552 the cleanup pass treated "no added frames" as a hard rule and declined is_title_piece for a single frame on comma credential runs: 24 ns a parse, no guard moved.

What it says.

• The scale of one frame, measured 2026-09-28 on py3.11: a trivial call costs about 13 ns, against 40–80 µs for a whole parse, so one frame per parse is roughly 0.03% of it.
• Four questions, in order:
  1. Does the cost multiply with the input?
  2. How many names pay it?
  3. Does it move an enforced guard (the ±2% _CALL_BASELINE band, or a scaling ratio)?
  4. What does the inline alternative cost to maintain? A drifting copy of a predicate is mechanisms.md#ONE-PREDICATE-PER-QUESTION; a narrower question stated exactly is not a copy.
• The converse from Parsing slows quadratically for a long given part after a family comma (Doe, Jane …) #553: the frame counter cannot see C-level work such as x in some_list, so "no new frames" is not "no new cost".

Docs only; the doc tests pass.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.85%. Comparing base (003b496) to head (3068de8).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #556   +/-   ##
=======================================
  Coverage   98.85%   98.85%           
=======================================
  Files          45       45           
  Lines        3934     3934           
=======================================
  Hits         3889     3889           
  Misses         45       45           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.