[APM/ML] Anomaly detection on error rate

# Summary

We already have an anomaly detection job available for transaction duration and with the newly added error rate chart for errors, it might be worth investigating setting up another anomaly detection job for error rate.

# Open questions

- The error rate is currently being calculated by dividing the number of transactions by the number of errors in a given time period. Something that is done in the UI. How is it possible for ML to create a job configuration that looks at this when it's in two different indices (transactions and errors)?
- How will the UX change? Will we simply add another job to the anomaly detection "bundle" when the user wants to set it up for their selected service?

cc @blaklaybul @nehaduggal @sqren @cauemarcondes