populate_db: Generate topics using config fixture.

Instread of using stream_name + Intergers as topics, we now
generate topics using pos in `config.generate_data.json`.

This helps us create and test more realistic topics.

zerver/lib/generate_test_data.py

@@ -12,6 +12,29 @@ def load_config() -> Dict[str, Any]:
 
     return config
 
+def generate_topics(num_topics: int) -> List[str]:
+    config = load_config()["gen_fodder"]
+
+    topics = []
+    # Make single word topics account for 30% of total topics.
+    # Single word topics are most common, thus
+    # it is important we test on it.
+    num_single_word_topics = num_topics // 3
+    for _ in itertools.repeat(None, num_single_word_topics):
+        topics.append(random.choice(config["nouns"]))
+
+    sentence = ["adjectives", "nouns", "connectors", "verbs", "adverbs"]
+    for pos in sentence:
+        # Add an empty string so that we can generate variable length topics.
+        config[pos].append("")
+
+    for _ in itertools.repeat(None, num_topics - num_single_word_topics):
+        generated_topic = [random.choice(config[pos]) for pos in sentence]
+        topic = " ".join(filter(None, generated_topic))
+        topics.append(topic)
+
+    return topics
+
 def load_generators(config: Dict[str, Any]) -> Dict[str, Any]:
 
     results = {}