fix(migrate): count only written sends as senders, note UI-posted specializations, keep a misrouted send's argument note, name only written clock waits

[devin-ai-integration]

What and why

Follow-up to #791, addressing the four post-merge review findings on its ledger logic. All four are generic to the migrator's report; no notation changes (the TMT export migrates to a byte-identical TMT.sysml, -validate still 0 errors; only the order of two joined notes moves).

1. A refused send no longer counts as a sender. recordSender indexed every SendSignalAction/BroadcastSignalAction during prepare, including one sendRefusal later writes as a placeholder (absent/dry/misfit argument), so an accept of a signal only the tool's UI prototype posts went un-noted when such a placeholder existed. Now sendSignal marks senders.sent[sig] only after it emits send new …;, and the UI-only notes are deferred: uiAccept records each accept (transition trigger, accept action) during writing, and uiOnlyAccepts — run after placeholderEnds, when every send is decided — annotates them through the report's pending-note path, so timing and verdict are unchanged for the cases already covered.

2. A button posting a specialization raises the note. unsentNote only looked up buttons[sig]; a button posting Long Pause :> Pause left an accept of Pause unnoted although the accept takes the specialization. buttonsPosting now counts buttons posting sig or a signal inheriting from it (m.inherits, the same test sentByAction already used for sends) and the note names the specializations:
   no send action of the document sends 'Pause' or a signal specializing it ('Long Pause'), which only 1 button ….

3. A send on a port that is not the sender's keeps its argument note. Both invalid-port branches of sendSignal assigned note = …, overwriting the argument note signalArguments had produced; they now joinNotes, like the target-pin branch already did.

4. clockWaitIn names only waits the writer writes. It reported a duration constraint as a wait whenever it had an interval (waitBound), and a time-event accept whenever the trigger named a TimeEvent, and followed called behaviors/methods regardless of whether the callee is written. An entry/exit/effect whose only "wait" the writer leaves as a placeholder (interval open above, time it cannot spell, callee without a v2 declaration) was therefore noted as stopping a run at a wait that does not exist. The duration decision is now factored into migration.waitOf, shared by activity.waitFor (the writer) and clockWaitIn; acceptsTime asks acceptClause (pure) whether the first trigger is writable, with the same scope the writer uses (n.Parent is the activity or structured node whose activity.act writes it); and a call is followed to its callee only when the writer emits it as an invocation: clockWaitIn links the caller's graph as the writer does (scanGraph: newActivity → link/resolveData/deaden, without writing) and activity.invokes applies the writer's own rules — a call refusal refuses (unbound context, undeclared callee, no operation) is a placeholder, a library primitive is no call, an operation written as a usage whose target pin resolves no receiver is written as the empty step; a node whose pin nothing fills is skipped, as the writer writes it without its wait.

Also carries the one-line shaped(mult, r, true, false) repair from #800, which develop needs to build; it drops out of this diff when #800 lands.

Specification basis

No runtime behaviour or compliance row moves; this is the migrator's ledger. UML 2.5.1 §13.3.3 (an accept of a signal takes instances of its specializations) is why a specialized posting counts for an accept of the general signal, on the UI side as it already did on the send side (docs/reference/sysml-v1-migration.md, SignalEvent and instant-wait rows updated).

How it was verified

• tests/migrate/ui_stimuli.xmi + TestSignalsOnlyTheToolsUIPostsAreLedgered: added Pause/Long Pause :> Pause with a button posting Long Pause (finding 2), and Resume with a required delay attribute, a send of it without an argument (a placeholder) and a button posting it (finding 1); _trPause and _trResume are noted, _sendResume carries its own argument note.
• tests/migrate/instant_waits.xmi + TestInstantBehaviorsWaitingForTheClockAreLedgered: added transition effect Send Ack Hold calling Hold (duration constraint with only a min → unmapped) and accepting an absolute time event whose time names nothing (→ placeholder); the effect gets no "waits for the clock" note while the existing four waits still do (finding 4); added Send Ack Settle, calling Relay::Settle (accepts a time event, reads gain through its context) from a Stimulus that cannot bind it — a placeholder — and Send Ack Pause, calling Delay through a target pin nothing feeds — never fires; neither is noted as waiting. Checked the never-fires shape against the runtime: an unconnected action call : Delay; beside first start then final; action final terminate; completes at clock 0.0, the connected one at 2.0.
• TestSendOnAnotherObjectsPortKeepsItsArgumentNote (new, in signal_absent_argument_test.go): a send with a surplus argument pin on another block's port; the note joins both reasons (finding 3).
• Full gates on the branch with the four corpora downloaded and required: go build ./... && go vet ./... && gofmt -l . clean; go test -count=1 ./... all packages ok (incl. tests/corpus, tests/export, tests/identity, tests/model, tests/migrate); python3 scripts/changelog.py check ok.
• TMT (TMT-2024x.mdzip + MTIP layout): migrated notation byte-identical to develop+fix(migrate): pass the uniqueness argument the read result pin's shaped multiplicity takes #800; -validate 0 errors; migration report differs only in the order of the two joined notes on the PEAS Abort and MBClock tick triggers (the UI-only note is now appended after the port note).

Checklist

• make test and make lint pass locally
• Tests added or updated for the change
• Documentation extended where it already covers the surface (see CONTRIBUTING.md)
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (compliance rows need nothing: the census is counted at docs build)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring