Edit tools' diffs are a mess in line ordering.

### Describe the bug

When copilot cli asks me for approval on edits, it very often makes an absolute mess on line ordering, making the review nearly impossible. In the example below you can see the line numbers make absolutely no sense.

```
───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ Edit file │
│ ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────── │
│ settler.py +44 -34 │
│ /my_service │
│
463    │
│ 464    │
│ 465 +   async def handle_settlement_sqs_message(msg_body: str) -> None: │
│ 466 +    await on_community_bucket_expired(msg_body) │
│ 467 +    │
│ 468 +    │
│ 469   class MeasurementHandler: │
│ 470    def __init__(self) -> None: │
│ 471    self._task: asyncio.Task[None] | None = None │
│ 472    │
│ 473    async def start(self) -> None: │
│ 458 -   """Schedule the Redis expiry listener task. Call from within an async context (e.g. FastAPI │
│ lifespan).""" │
│ 474 +   """Schedule the settlement scanner task. Call from within an async context (e.g. FastAPI lifespan).""" │
│ │
│ 475    log = logging.getLogger(__name__) │
│ 476    log.info("Starting S2PowerMeasurementHandler") │
│ 477    self._task = asyncio.create_task( │
│ 474   """Schedule the key scans""" │
│ 475    log = logging.getLogger(__name__) │
│ 476    log.info("Starting MeasurementHandler") │
│ 461 -    self._task = asyncio.create_task( │
│ 462 -    start_expiry_listener( │
│ 463 -    { │
│ 464 -    "my_service:measurements:*:last_message": on_community_bucket_expired │
│ 465 -    } │
│ 466 -    ) │
│ 467 -    ) │
│ 477 +    self._task = asyncio.create_task(settlement_scanner()) │
│ 478    │
│ 479    def task_done_callback(t: asyncio.Task[None]) -> None: │
│ 480    if t.cancelled(): │
│ 471 -    log.info("Expiry listener task was cancelled") │
│ 481 +    log.info("Settlement scanner task was cancelled") │
│ 482    elif t.exception(): │
│ 483    log.exception("Expiry listener task failed", exc_info=t.exception()) │
│ 484    else: │
│ 480    if t.cancelled(): │
│ 481    log.info("Expiry listener task was cancelled") │
│ 482    elif t.exception(): │
│ 473 -    log.exception("Expiry listener task failed", exc_info=t.exception()) │
│ 483 +    log.exception("Settlement scanner task failed", exc_info=t.exception()) │
│ 484    else: │
│ 475 -    log.info("Expiry listener task completed normally") │
│ 485 +    log.info("Settlement scanner task completed normally") │
│ 486    │
│ 487    self._task.add_done_callback(task_done_callback) │
│ 488    │
│ 487    self._task.add_done_callback(task_done_callback) │
│ 488    │
│ 489    async def stop(self) -> None: │
│ 480 -    """Cancel the Redis expiry listener task and wait for it to finish.""" │
│ 490 +    """Cancel the settlement scanner task and wait for it to finish.""" │
│ 491    if self._task is not None: │
│ 492    self._task.cancel() │
│ 493    try: │
│
```

### Affected version

GitHub Copilot CLI 1.0.45.

### Steps to reproduce the behavior

I run copilot on the cli and ask it for confirmation of every edit so i review and guide it as its writing code (rather than a single large diff)

### Expected behavior

_No response_

### Additional context

_No response_

Non sono indicati file, test o punti di ingresso. Riproduci il flusso di approvazione in GitHub Copilot CLI 1.0.45 con modifiche incrementali, quindi traccia il percorso di rendering di edit-diff; il lavoro è completato quando i numeri di riga visualizzati e le righe modificate seguono l’ordine effettivo del file e il comportamento è coperto da un controllo di regressione.

Scritto dal modello di indicizzazione a partire dal testo della issue.