fix(vector): don't inflate qdrant-error metric on embed drops (#893 r3)

Round-3 review on PR #893:
- record_qdrant_operation("upsert","error") now fires only when the exhausted
  retry was actually a Qdrant failure (reason=="qdrant"); an embed/connection
  failure exhausts retries before Qdrant is called, so attributing it to
  mcp_qdrant_operations_total{error} inflated that signal. The cause is still
  captured by record_ingest_dropped.
- Add test_mistral_embed_retries_on_5xx: exercises the full Mistral retry path
  (5xx SDKError then success), not just the predicate.
- Add test_generate_does_not_retry_on_bad_request: generate() fast-fails on a
  permanent 4xx.
- Move astrolabe_vector_ingest_dropped_total's definition into the astrolabe_
  pipeline-metrics block (was in the mcp_ section).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

tests/unit/providers/test_openai.py

@@ -457,3 +457,25 @@ async def test_generate_retries_on_connection_error(mock_openai_client, monkeypa
     text = await provider.generate("prompt")
     assert text == "Generated response"
     assert create.await_count == 2
+
+
+@pytest.mark.unit
+async def test_generate_does_not_retry_on_bad_request(mock_openai_client, monkeypatch):
+    """generate() fast-fails (no retry) on a permanent 4xx."""
+    import httpx
+    from openai import BadRequestError
+
+    from nextcloud_mcp_server.providers import _retry
+
+    monkeypatch.setattr(_retry.anyio, "sleep", AsyncMock(return_value=None))
+
+    err = BadRequestError(
+        "bad", response=httpx.Response(400, request=_req()), body=None
+    )
+    create = AsyncMock(side_effect=err)
+    mock_openai_client.chat.completions.create = create
+    provider = OpenAIProvider(api_key="test-key", generation_model="gpt-4o-mini")
+
+    with pytest.raises(BadRequestError):
+        await provider.generate("prompt")
+    assert create.await_count == 1  # no retry on a permanent 4xx