[ML] Optimize the job calendar update procedure #137872
Description
In #136886, we implemented a change that updates each job one-by-one. Each update requires a query to the calendar index from the ML node to retrieve the current calendar events. This means potentially thousands of queries within a short period of time.
We ca optimize this procedure by grouping the updates by nodes that contain the corresponding job processes. Then each ML node can query the calendar index only once and update all job processes it contains.
cc @davidkyle