Wire aggregation job failures into per-task counters

Janus tracks a variety of per-task metrics to count the occurrence of things like report uploads being rejected due to HPKE decryption failures. But if a report gets rejected by the Helper, the Leader's per-task counters won't reflect these failures. In particular, if Clients are configured to use the wrong HPKE keys for the Helper, then reports will fail to be aggregated but no information about this will be available via the Leader-side task counters. This can make it harder for someone deploying a task and clients to debug, as they may not have access to the Helper's per-task counters, Prometheus metrics, or logs.

We could expose some new per-task counters, clearly labeled by aggregator, and increment them as the Leader walks the prepare_resps in an AggregationJobResp.

Start by tracing the Leader's handling of AggregationJobResp and its prepare_resps entries. Identify the existing per-task counter definitions and increment path, then add clearly aggregator-labeled counters for Helper-side report rejections and verify that failures such as HPKE decryption errors are reflected in Leader-side task metrics.