feat: fast_search in Python and Node (#1623)

Sometimes it is acceptable to users to only search indexed data and skip
and new un-indexed data. For example, if un-indexed data will be shortly
indexed and they don't mind the delay. In these cases, we can save a lot
of CPU time in search, and provide better latency. Users can activate
this on queries using `fast_search()`.

python/python/tests/test_query.py

@@ -17,6 +17,7 @@ from typing import Optional
 
 import lance
 import lancedb
+from lancedb.index import IvfPq
 import numpy as np
 import pandas.testing as tm
 import pyarrow as pa
@@ -358,6 +359,25 @@ async def test_query_to_pandas_async(table_async: AsyncTable):
     assert df.shape == (0, 4)
 
 
+@pytest.mark.asyncio
+async def test_fast_search_async(tmp_path):
+    db = await lancedb.connect_async(tmp_path)
+    vectors = pa.FixedShapeTensorArray.from_numpy_ndarray(
+        np.random.rand(256, 32)
+    ).storage
+    table = await db.create_table("test", pa.table({"vector": vectors}))
+    await table.create_index(
+        "vector", config=IvfPq(num_partitions=1, num_sub_vectors=1)
+    )
+    await table.add(pa.table({"vector": vectors}))
+
+    q = [1.0] * 32
+    plan = await table.query().nearest_to(q).explain_plan(True)
+    assert "LanceScan" in plan
+    plan = await table.query().nearest_to(q).fast_search().explain_plan(True)
+    assert "LanceScan" not in plan
+
+
 def test_explain_plan(table):
     q = LanceVectorQueryBuilder(table, [0, 0], "vector")
     plan = q.explain_plan(verbose=True)