[prometheus check] Actually use provided "metrics_mapper" (#3446)

The current implementation of the Prometheus check ignores completely the values of `metrics_mapper`, because the code is pretty much building it using the original metric names.

Instead, we should use the mappings provided by the user in the `metrics_mapper` dictionary. So given an example config like this:

```yml
    metrics:
      - cp_kafka_server_replicamanager_underreplicatedpartitions
    type_overrides:
      cp_kafka_server_replicamanager_underreplicatedpartitions: gauge
    metrics_mapper:
      cp_kafka_server_replicamanager_underreplicatedpartitions: server.replica_manager.under_replicated_partitions
```

We should emit metrics named `server.replica_manager.under_replicated_partitions` instead of `cp_kafka_server_replicamanager_underreplicatedpartitions` which is what the agent does currently.

datadog_checks_base/datadog_checks/base/checks/prometheus/base_check.py

@@ -146,6 +146,8 @@ def get_scraper(self, instance):
                 metrics_mapper[metric] = metric
             else:
                 metrics_mapper.update(metric)
+        # update metrics mapper with values provided by user
+        metrics_mapper.update(instance.get("metrics_mapper", {}))
 
         scraper.metrics_mapper = metrics_mapper
         scraper.labels_mapper = default_instance.get("labels_mapper", {})