[#26119] DocDB: Fix PgIndexBackfillTest.VectorIndex flakiness

Summary:
Even in single dimension HNSW could produce incomplete result.
And we cannot use it for consistency testing.
Relaxed requirements for query based check.
It is not a problem, since we already use VerifyVectorIndexes for consistency check.

Also ensured that intents applied before VerifyVectorIndexes.
Jira: DB-15446

Test Plan: Jenkins

Reviewers: arybochkin

Reviewed By: arybochkin

Subscribers: ybase, yql

Tags: #jenkins-ready

Differential Revision: https://phorge.dev.yugabyte.com/D42237

java/yb-pgsql/src/test/java/org/yb/pgsql/TestYbQueryDiagnostics.java

@@ -332,9 +332,6 @@ private Path runQueryDiagnostics(Statement statement, String queryId,
         return Paths.get(resultSet.getString("yb_query_diagnostics"));
     }
 
-    private void waitForBundleCompletion(String queryId, Statement statement) throws Exception {
-        waitForBundleCompletion(queryId, statement, 0);
-    }
     /*
      * Waits for the bundle to complete by checking the yb_query_diagnostics_status view.
      */

src/yb/yql/pgwrapper/pg_index_backfill-test.cc

@@ -2669,11 +2669,18 @@ struct VectorIndexWriter {
   }
 
   void Verify(PGConn& conn) {
+    int num_bad_results = 0;
     for (int i = 2; i < counter.load(); ++i) {
       auto rows = ASSERT_RESULT(conn.FetchAllAsString(Format(
           "SELECT id FROM test ORDER BY embedding <-> '[$0]' LIMIT 3", i * 1.0 - 0.01)));
-      ASSERT_EQ(rows, Format("$0; $1; $2", i, i - 1, i + 1));
+      auto expected = Format("$0; $1; $2", i, i - 1, i + 1);
+      if (rows != expected) {
+        LOG(INFO) << "Bad result: " << rows << " vs " << expected;
+        ++num_bad_results;
+      }
     }
+    // Expect recall 98% or better.
+    ASSERT_LE(num_bad_results, counter.load() / 50);
   }
 };
 
@@ -2710,6 +2717,10 @@ TEST_F(PgIndexBackfillTest, VectorIndex) {
   LOG(INFO) << "Max time without inserts: " << writer.max_time_without_inserts;
   ASSERT_LT(writer.max_time_without_inserts, 1s * kBackfillSleepSec);
   SCOPED_TRACE(Format("Total rows: $0", writer.counter.load()));
+
+  // VerifyVectorIndexes does not take intents into account, so could produce false failure.
+  ASSERT_OK(cluster_->WaitForAllIntentsApplied(30s * kTimeMultiplier));
+
   for (size_t i = 0; i != cluster_->num_tablet_servers(); ++i) {
     tserver::VerifyVectorIndexesRequestPB req;
     tserver::VerifyVectorIndexesResponsePB resp;