txns: fix list txns failure when it's used before find_coordinator

find_coordinator creates txn coordinator's topic when it's used for
the first time. list txns api expects the topic to be already created
so it used to fail when the api was invoked before find_coordinator.
Fixed the problem by creating the topic on the first list txns call.

fixes #11947

(cherry picked from commit 85ad993)

src/v/cluster/tx_gateway_frontend.cc

@@ -2953,6 +2953,12 @@ ss::future<> tx_gateway_frontend::do_expire_old_tx(
 
 ss::future<tx_gateway_frontend::return_all_txs_res>
 tx_gateway_frontend::get_all_transactions() {
+    if (!_metadata_cache.local().contains(model::tx_manager_nt)) {
+        if (!co_await try_create_tx_topic()) {
+            co_return tx_errc::unknown_server_error;
+        }
+    }
+
     auto shard = _shard_table.local().shard_for(model::tx_manager_ntp);
 
     if (!shard.has_value()) {

tests/rptest/tests/transaction_kafka_api_test.py

@@ -28,7 +28,7 @@ def __init__(self, test_context):
                                  "tx_timeout_delay_ms": 10000000,
                                  "abort_timed_out_transactions_interval_ms":
                                  10000000,
-                                 'enable_leader_balancer': False
+                                 "enable_leader_balancer": False
                              })
 
         self.kafka_cli = KafkaCliTools(self.redpanda, "3.0.0")
@@ -106,6 +106,11 @@ def test_describe_transactions(self):
                 tpoic_partition = f"{topic}-{partition}"
                 assert tpoic_partition in expected_partitions
 
+    @cluster(num_nodes=3)
+    def test_empty_list_transactions(self):
+        txs_info = self.kafka_cli.list_transactions()
+        assert len(txs_info) == 0
+
     @cluster(num_nodes=3)
     def test_list_transactions(self):
         producer1 = ck.Producer({