Add P104 flag 0 coefficient transport audit

[1337isnot1337]

Audit the exact native coefficient for P104 joint flag 0

Summary

Adds a bounded exact row audit for the first flag in PR #2981's 208-flag
collection: source X0, target Y34, basis head 232.

The pinned doubled E8 roots have raw dot product 4 and squared norm 8. The
native channel therefore contributes +1/2 to this off-diagonal entry, while
the complementary I-G/2 term requires -1/2. A one-hot x0=z replay through
the pinned 4,058-event word follows the target coefficient through events
2815 (+z/2), 2998 (-z/2), and 3369 (+z/2), finishing at Y34 = z/2.

This complements PR #3242's event-order scan and the pooled fixed-cover support
reviews in PRs #3259, #3803, and #3810. It checks an actual native coefficient
row for one flag; the three additions do not implement the side correction.

Reproduction

python3 -I -B research/p104-flag0-coefficient-transport/verify.py
python3 scripts/research_package.py --package research/p104-flag0-coefficient-transport --check-only
python3 scripts/check_research_commit.py --package research/p104-flag0-coefficient-transport

Local exact replay passed on CPython 3.14.8. The registered Python 3.12 replay
remains pending because that interpreter is unavailable in the local
environment.

Scope and limits

No correction program, arbitrary-dirty restoration, correction-frame
availability, full bill, supplier admission, new kappa, or selected-frontier
change is claimed. The one-hot fixture sets dirty coefficients to zero and uses
the coefficient-row invariant in the reviewed native-channel proof. It does not
re-run the native Clifford/all-column checker.

Credit Nick Petalas / PR #2981 for the static flag; Rohan Arun / PR #1972 and
Chafik Boukhalfa / PR #1991 for the channel; and jwr404 / PR #3242 for the
event chronology. No independent human or formal-kernel review is claimed.

Current overlap

Open PR #3882 submits a P104 correction-tail construction. This package checks one native coefficient row for flag 0 and does not construct a correction program.

Publication validation

python3.12 scripts/research_package.py --package research/p104-flag0-coefficient-transport passed on CPython 3.12.3, and python3.12 scripts/check_research_commit.py --package research/p104-flag0-coefficient-transport --check-source-history passed. The registered replay left the worktree clean. This supersedes any earlier pending note above for the registered package replay; separate full selected cold-replay limits remain as stated.