porch records only FAILING consultation rounds in status.yaml history, so a phase that passes first time leaves no record

[pseudoseed]

What happens

On spec 250, 20 implementation consultation rounds ran (2 lanes each, 43 lane files on disk).
status.yaml history records 9.

Recorded:

phase_1 iter1   phase_2 iter1   phase_3 iter1
phase_4 iter1   phase_4 iter2   phase_5 iter1
phase_6 iter1   phase_10 iter1  phase_11 iter1

The pattern is exact

Cross-referencing every recorded round against its verdicts, a round is recorded if and only if
at least one lane did not approve.

• phases 7, 8 and 9 are absent entirely. They are the three phases where both lanes approved on
  round 1. They have lane files on disk and no history entry at all.
• Every terminal (approving) round is missing. phase_1 iter2, phase_2 iter2, phase_3 iter2,
  phase_4 iter3, phase_5 iter2, phase_6 iter2, phase_10 iter2, phase_11 iter2 — all APPROVE on both
  lanes, none recorded.
• phase_4 iter2 is recorded, and it is the one middle iteration that still carried a
  REQUEST_CHANGES (opencode).

There is no exception in either direction across 20 rounds.

Why it matters

history is the only machine-readable record of what review actually happened, and it currently
shows the rework and never the approval that ended it. Two concrete consequences:

1. A phase reviewed cleanly is indistinguishable from a phase never reviewed. Both have no
   entry. That is the "could not tell must not be spelled like no" rule broken in the protocol's
   own state file.
2. history understates review effort by more than half (9 of 20), and it understates it
   selectively — the record is biased toward phases that went badly.

Anyone auditing a project's review coverage from status.yaml will draw the wrong conclusion, and
the artifacts that would correct them (codev/projects/*/*.txt) are gitignored, so they do not
travel with the PR.

Expected

Every consultation round appends to history, whatever its verdicts. The approving round is the
most useful one to record: it is the one that says the phase was cleared, and by whom.

Reproduce

Any SPIR project where a phase's consultation approves on the first round. Compare:

# rounds that ran
ls codev/projects/<id>-*/ | grep '\.txt$' \
  | sed 's/-\(claude\|opencode\|codex\|gemini\)\.txt$//' | sort -u | wc -l
# rounds recorded
grep -c 'plan_phase:' codev/projects/<id>-*/status.yaml

Found by

Spec 250, review phase — raised by the claude lane's PR review of #266 and confirmed against the
artifacts. status.yaml was not hand-edited.