rptest: make target_min_capacity_wanted more deterministic

Introduce linger.ms parameter to try and make batching in the test more
deterministic when test is run on systems with different
performance/load.

(cherry picked from commit 98b6add)

tests/rptest/clients/kafka_cli_tools.py

@@ -320,7 +320,8 @@ def produce(self,
                 record_size: int,
                 acks: int = -1,
                 throughput: int = -1,
-                batch_size: int = 81960):
+                batch_size: int = 81960,
+                linger_ms: int = 0):
         self._redpanda.logger.debug("Producing to topic: %s", topic)
         cmd = [self._script("kafka-producer-perf-test.sh")]
         cmd += ["--topic", topic]
@@ -329,9 +330,11 @@ def produce(self,
         cmd += ["--throughput", str(throughput)]
         cmd += [
             "--producer-props",
-            "acks=%d" % acks, "client.id=ducktape",
+            "acks=%d" % acks,
+            "client.id=ducktape",
             "batch.size=%d" % batch_size,
-            "bootstrap.servers=%s" % self._redpanda.brokers()
+            "bootstrap.servers=%s" % self._redpanda.brokers(),
+            "linger.ms=%d" % linger_ms,
         ]
         if self._command_config:
             cmd += ["--producer.config", self._command_config.name]

tests/rptest/tests/full_disk_test.py

@@ -377,7 +377,8 @@ def test_target_min_capacity_wanted_time_based(self):
                             30 * 1024,
                             1024,
                             throughput=500,
-                            acks=-1)
+                            acks=-1,
+                            linger_ms=50)
 
         node = self.redpanda.nodes[0]
         reported = admin.get_local_storage_usage(