test(node): cover foreign Float64 vector schema workflow (#3844)

## Summary

- add an end-to-end regression for schemas created by a different Apache
Arrow package instance
- cover seeded table creation, filtered scanning, and Float64 vector
search across Arrow 15–18

## Root cause

Apache Arrow's runtime identity checks historically rejected schemas
created by another installed Arrow instance, producing the constructor
failures reported in the issue. LanceDB's peer dependency and
foreign-schema sanitization now handle that boundary, but the complete
reported workflow was only covered by separate unit tests. This
regression keeps the repaired behavior protected end to end.

## Validation

- `pnpm exec jest --runInBand __test__/table.test.ts` (281 passed)
- `pnpm lint-ci`
- `pnpm build`
- `pnpm run docs`

Fixes #882

<!-- lance-gatekeeper-fix:v1 agent=43b19dea581cfbc83ee1e9ed21a335a6
generation=1 -->

Co-authored-by: Gatefixer <313497061+lancedb-gatefixer[bot]@users.noreply.github.com>

nodejs/__test__/table.test.ts

@@ -86,6 +86,44 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       await expect(table.countRows()).resolves.toBe(3);
     });
 
+    it("should support a foreign Float64 vector schema end to end", async () => {
+      const conn = await connect(tmpDir.name);
+      const schema = new arrow.Schema([
+        new arrow.Field("resource_id", new arrow.Int32(), false),
+        new arrow.Field(
+          "vector",
+          new arrow.FixedSizeList(
+            3,
+            new arrow.Field("value", new arrow.Float64(), true),
+          ),
+          false,
+        ),
+      ]);
+      const data = [
+        {
+          // biome-ignore lint/style/useNamingConvention: matches the reported schema
+          resource_id: 0,
+          vector: [0.1, 0.1, 0.1],
+        },
+      ];
+
+      const resources = await conn.createTable("resources", data, { schema });
+
+      const existing = await resources
+        .query()
+        .where("resource_id = 0")
+        .limit(1)
+        .toArray();
+      expect(existing).toHaveLength(1);
+
+      const matched = await resources
+        .search(Float64Array.from(data[0].vector))
+        .limit(1)
+        .toArray();
+      expect(matched).toHaveLength(1);
+      expect(matched[0]["resource_id"]).toBe(0);
+    });
+
     it("should support branches", async () => {
       await table.add([{ id: 1 }]);
       expect(await table.countRows()).toBe(1);