fix(providers): allow retry and fallback on stream stalled timeout

When a stream stalls mid-response, both the retry layer and
FallbackProvider blocked recovery because content had already been
emitted via on_content_delta. This left users with truncated replies
and no automatic recovery.

For error_kind="timeout" specifically:
- _run_with_retry now suppresses delta callbacks and retries the same
  model instead of returning immediately
- FallbackProvider now allows failover to a different model with
  delta callbacks suppressed

Non-timeout errors retain the original "skip retry/failover after
streamed content" behavior to avoid duplicate output.

tests/providers/test_provider_retry.py

@@ -163,6 +163,42 @@ async def test_chat_stream_with_retry_does_not_retry_after_emitting_content(monk
     assert delays == []
 
 
+@pytest.mark.asyncio
+async def test_chat_stream_with_retry_retries_timeout_after_emitting_content(monkeypatch) -> None:
+    first = LLMResponse(
+        content="Error calling LLM: stream stalled for more than 30 seconds",
+        finish_reason="error",
+        error_kind="timeout",
+    )
+    first._test_stream_delta = "partial"  # type: ignore[attr-defined]
+    provider = ScriptedProvider([
+        first,
+        LLMResponse(content="full retry response"),
+    ])
+    deltas: list[str] = []
+    delays: list[int] = []
+
+    async def _fake_sleep(delay: int) -> None:
+        delays.append(delay)
+
+    async def _on_delta(delta: str) -> None:
+        deltas.append(delta)
+
+    monkeypatch.setattr("nanobot.providers.base.asyncio.sleep", _fake_sleep)
+
+    response = await provider.chat_stream_with_retry(
+        messages=[{"role": "user", "content": "hello"}],
+        on_content_delta=_on_delta,
+    )
+
+    assert response.content == "full retry response"
+    assert response.finish_reason == "stop"
+    assert provider.calls == 2
+    assert deltas == ["partial"]
+    assert delays == [1]
+    assert provider.last_kwargs.get("on_content_delta") is None
+
+
 @pytest.mark.asyncio
 async def test_chat_with_retry_uses_provider_generation_defaults() -> None:
     """When callers omit generation params, provider.generation defaults are used."""