Ignore rows which are missing stats during daily aggregation (#5287)

There were a couple old versions of CF running in the
clusterfuzz-candidate2 bots. This PR ignores the stats which were
generated by those bots. It's unfortunate that the stats are missing
these, but IMO it's better to completely ignore these rows rather than
writing partial data. There were 10 of these instances running, which
resulted in some fuzzers having a fuzzing session unaccounted for.

Testing: Ran this SQL against the BigQuery table and verified it
correctly filters out the rows and resulted in the `testcases_generated
== testcases_executed`

### Rollout

we will have to rerun the cron job for the past month again after this
deploys to properly ignore those rows.

Author: dylanjew

src/clusterfuzz/_internal/cron/aggregate_fuzzer_stats.py

@@ -149,6 +149,8 @@ def _query_fuzzer_stats(fuzzer_name, project_id, target_date_str):
   dataset_id = fuzzer_stats.dataset_name(fuzzer_name)
   table_id = 'JobRun'
 
+  # We ignore rows where testcases_generated is NULL because those are rows
+  # created by an old revision of Clusterfuzz which are missing the durations.
   query = f"""
   SELECT
   '{fuzzer_name}' as fuzzer_name,
@@ -170,6 +172,7 @@ def _query_fuzzer_stats(fuzzer_name, project_id, target_date_str):
   FROM    `{project_id}.{dataset_id}.{table_id}`
   WHERE
     DATE(TIMESTAMP_SECONDS(CAST(timestamp AS INT64))) = '{target_date_str}'
+    AND testcases_generated IS NOT NULL
   GROUP BY
     date
   """