docs(plan): record r27 - the flag and settings probes, the fixes and the cleanup batch - #64

Summary

Records r27: the five PRs of 2026-10-30 (#57, #60, #61, #62, #63) and what the user decided. Documents only. The text was written when the five PRs were open, so the plan says "done, not merged" and gives the PR numbers.

execution log   + section 10.30 (r27): measure (#57), fix (#60), picker cutoff (#61),
                  dead modules (#62), test-support crate (#63)
                  what is left open; what is not verified; merge order and conflicts
                  contents list, section 10 title and intro, revision log
plan            header, "current state", phase D row, "not in V0.6", V0.7 known issues
                next steps (3 bullets), "not verified yet" (flags, settings)
                decisions 8, 9, 10, 14 and the lists above them
                section 7: 1.P6, D.5, D.8
                section 10 index row, revision summary, link definitions

What the plan says the user decided (2026-10-30), and no more:

• Decisions 8 and 14: "measure, then change as the plan leans" (hide the dead settings, refuse the dead flags with a reason, Python unchanged; open PRs, do not merge).
• The cleanup batch (decision 9, the two open items of decision 10, D.8, D.5): "also OK to do".
• The user did not choose the method for decision 9. The plan says that the first option of the tendency was done and "configurable" was not.

docs/baselines/tui.md is not changed. Its numbers come with CI runs on main, so they are measured after the merge. (The body of #60 said this PR would update it; that sentence is corrected.)

Evidence

• Before: the plan lists decisions 8, 9, 14 and two items of decision 10 as open, and says the flags and the settings page are not verified.
  After: they are "done" with the PR numbers, each marked "not merged". "Not verified yet" says what the probes measured (Linux only, debug build for the PTY regression) and what they did not.
• Link check (scratch script), also after the rebase on the stack: every [text][ref] of the plan has a definition (pr-57 to pr-63 and l10-30 are new); every ](#anchor) of both files has a heading; the new anchors follow the GitHub rules. gh api markdown (gfm) shows no stray **.
• The numbers in section 10.30 come from the bodies and the runs of the five PRs. CI of the five PRs when this was written: all green (Ruff and pytest 3.11 to 3.13 for each; Gates + cargo check and cargo nextest too for fix(tui): hide eight dead settings and refuse flags the Python agent does not back #60 to refactor(tui): remove the unused half of pi-test-support (plan D.5) #63).
• ruff check and ruff format --check pass. No code changed, so no test ran.

Merge Danger

Door: two-way.

Blast Radius: documentation (two files under docs/PLAN/).

Round numbers. This PR is r27 / section 10.30 and sits above #63 in the stack (#58, #59, #56, #57, #60, #61, #62, #63, this PR, #65). PR #58 and PR #59 (the other session) hold r24 / 10.27 and r25 / 10.28. PR #56 was r24 too; it is r26 / 10.29 now. No number is used twice. The rebase changed the numbers and the lists of rounds in the plan and the log. It did not change the r27 text, except the sentences about the round numbers.

Stack. All ten PRs are merged by one gh stack merge. From that moment the plan lines that say "not merged" are out of date. A follow-up docs PR (r28) updates them and records the merge (commits, order, CI on main, the numbers of docs/baselines/tui.md).

The code PRs. This PR says what they do. It does not change them. If one of the five is closed instead of merged, the plan lines for it must change.