[yugabyte#27576] Docdb: Reduce PgIndexBackfill test PhantomIdxEntry/1 flakiness

Summary:
PgIndexBackfillReadCommittedBlockIndislive.PhantomIdxEntry/1 is flaky and fails about 7/30 times sporadically.

This is due to a DML running right after Create Index, that may run into a "schema version mismatch".
Add a retry around the DML to make the test less flaky.
Jira: DB-17139

Test Plan:
Jenkins: test regex: PgIndexBackfillReadCommittedBlockIndislive.PhantomIdxEntry/1

yb_build.sh release  --cxx-test pg_index_backfill-test --gtest_filter PgIndexBackfillReadCommittedBlockIndislive.PhantomIdxEntry/1 --clang19 -n 30 --tp 3

Before this change, on master: 7 / 30 failures
Before yugabyte@e01d18c: 7 / 30 failures, so does not appear to be caused by that diff.
After this diff: 0 / 30

Reviewers: jason, hsunder

Reviewed By: jason, hsunder

Subscribers: ybase, yql

Differential Revision: https://phorge.dev.yugabyte.com/D44674

Author: amitanandaiyer

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

@@ -2384,7 +2384,17 @@ TEST_P(PgIndexBackfillReadCommittedBlockIndislive, PhantomIdxEntry) {
   ASSERT_OK(cluster_->SetFlagOnTServers("ysql_yb_test_block_index_phase", "indisready"));
   ASSERT_OK(WaitForIndexStateFlags(index_live_flags, kIndexName));
   LOG(INFO) << "Update record by newer txn";
-  ASSERT_OK(conn_->ExecuteFormat("UPDATE $0 SET t = 'b' WHERE i = $1", kTableName, 2));
+  ASSERT_OK(LoggedWaitFor(
+      [this]() -> Result<bool> {
+        auto s = conn_->ExecuteFormat("UPDATE $0 SET t = 'b' WHERE i = $1", kTableName, 2);
+        if (s.ok()) {
+          return true;
+        }
+        SCHECK_STR_CONTAINS(s.message().ToBuffer(), "schema version mismatch");
+        return false;
+      },
+      RegularBuildVsSanitizers(10s, 30s),
+      "wait for DML to succeed, ignoring potential schema version mismatch"));
   LOG(INFO) << "Update record by older txn";
   ASSERT_OK(other_conn.ExecuteFormat("UPDATE $0 SET t = 'c' WHERE i = $1", kTableName, 2));
   ASSERT_OK(other_conn.Execute("COMMIT"));