[ML] Operational alerting rule type for Anomaly detection jobs

[darnautov]

During the lifecycle of the anomaly detection job, certain issues may occur that eventually prevent it from detecting anomalies and may result in missing alerts about them. We should introduce a new alerting rule type to monitor the job's health and notify the user accordingly.

The list of tests to perform:

• Datafeed is not started (7.15 - [ML] Alerting rule for Anomaly Detection jobs monitoring #106084)
• Job that is opened has reached SOFT_LIMIT or HARD_LIMIT model memory limit (7.15 - [ML] Model memory state test for Anomaly Detection jobs health rule type #106763)
• Delayed data has occurred (7.15 - [ML] Delayed data test for Anomaly Detection jobs health rule type #107183)
• Job is running behind real-time
• There are errors in job messages (7.15 - [ML] Check for error messages in the Anomaly Detection jobs health rule type #108701)

[elasticmachine]

Pinging @elastic/ml-ui (:ml)