Preserve token-union proposals in Go #900

Problem

The classifier at packages/agents/source/skills/resolve-rebase-conflicts/shared/scripts/conflict_shape.py:140 uses line alignment to find one-line replacements extended with tokens on both sides. Its output is a proposal, not a mechanical resolution. A generic diff algorithm can align repeated lines differently and offer a different proposal.

Part of #588 and #326.

Acceptance criteria

1. AC1: Given both sides extend an aligned single ancestor line without removing its tokens, when proposal analysis runs, then it identifies the same eligible lines as the frozen classifier.
2. AC2: Given multiline replacements, ancestor-token deletion, absent versions or identical side token sequences, when proposal analysis runs, then it offers no proposal for those cases.
3. AC3: Given an eligible line, when its proposal is printed, then ancestor tokens precede new base-side and replayed-side tokens in the frozen order.
4. AC4: Given indentation, duplicate tokens or byte whitespace, when its proposal is printed, then formatting matches the frozen output.
5. AC5: Given repeated lines and long repeated sequences, when proposal analysis runs, then alignment and line numbering match the frozen classifier.
6. AC6: Given a proposal, when it is reported, then it explicitly leaves ordering to the operator.
7. AC7: Non-functional: proposal analysis never edits or stages a resolution.
8. AC8: Non-functional: native proposal cases execute without skips on macOS, Linux and Windows.

Scope and non-goals

Scope: token eligibility, line alignment and proposal rendering.

Non-goal: automatic application, operator confirmation machinery, a new diff library, or changes to sorted-union eligibility. Keep shell and coupled Python frozen through N-2, with no new shell logic. Do not switch callers or retire tests in this child.

Approach

Port the observable alignment semantics of the existing Python SequenceMatcher call, including its repeated-line heuristic, rather than substituting Git diff. Render the evidence at packages/agents/source/skills/resolve-rebase-conflicts/shared/scripts/conflict_shape.py:204. The vendored matcher at vendor/github.com/stretchr/testify/internal/difflib/difflib.go:83 is internal to its dependency and is not an importable production API. Budget any local matcher code or newly vendored source in full; do not exclude it from the size. Keep this bounded to the alignment needed for proposals.

Dependencies

Blocked by: #899 (for shared stage interpretation), #753 (before registration).

Size

M. Forecast 720 code lines: 330 alignment/proposal implementation, 350 tests and fixtures, 40 integration. Baseline: the token sections of the 212-code-line helper at packages/agents/source/skills/resolve-rebase-conflicts/shared/scripts/conflict_shape.py:140. Prose allowance: 180 lines.

scc could not run in the isolated checkout. Measured baselines count non-blank lines other than full-line # comments, including Python docstrings and embedded code conservatively. Forecasts include implementation, tests, fixtures and registration; they are estimates, not measured future Go. Generated output is excluded only under packages/agents/claude/.apm/skills/, packages/agents/codex/.apm/skills/, .claude/skills/, .agents/skills/ and tokens.json. The source under packages/agents/source/ and copies anywhere else count. The final diff must be at most 800 code lines and 2,000 prose lines, with no exception; split before implementation if either forecast grows beyond its limit.

Verify plan

• AC1, AC2, AC3, AC4: compare exact proposals against the frozen helper for eligible and ineligible replacements, token order, indentation and byte whitespace.
• AC5: fixtures cover repeated lines and sequences on either side of the repeated-line heuristic boundary, with expected aligned line numbers.
• AC6, AC7: assert the ordering instruction is present and repository/index snapshots are unchanged.
• AC8: inspect native CI results for all promised platforms.
• Definition of done: D1 each AC is verified at the merged head, with the commands and results cited in the pull request's Verification. D2 an independent review approves the exact merged head, and every finding has a recorded disposition. D3 required checks pass with no gate loosened. D4 the leak scans the repository specifies are clean, and every post made for this work passes the per-post leak review before it is posted. D5 every commit is signed. D6 the documentation matches the change or the Verification says why there is no impact. D7 every commit has an Assisted-by trailer and the pull request has the provenance:agent-assisted label. D8 the pull request stays within the size limits. D9 follow-ups are filed as issues and listed under Related. D10 the merge closes the issue through a closing reference, the board shows Done, and the final Status: comment names the merge commit.

Risks

Different alignment may suggest the wrong combined line. Copying an entire diff dependency can exceed the limit; split alignment into a prerequisite child before coding if the forecast does not fit. Undo: revert this unused Go implementation.

Operator authority

A choice the operator reserved: token ordering remains a runtime operator decision, as packages/agents/source/skills/resolve-rebase-conflicts/shared/scripts/conflict_shape.py:215 requires; this port grants no automatic application authority.