Summary

close #141765

This is a follow-up to #139349 that improves search-session saved object update code. This is what was done and why:

1. Since [Search/Search Sessions] Stabilization Stage I #139349 we explicitly create a search-session object, we can simplify old create/update logic, by splitting create and update (remove updateOrCreate method)
2. As we split create and update and as SavedObjects update can cause version conflict errors even when not using optimistic concurrency control #126240 was fixed, we remove custom retry on conflict logic. Now we can rely on elasticsearch's retryOnConflict.
3. Change default sessionConfig.maxUpdateRetries from 3 to 10. This is done because the old "retry" default was actually 3 on Kibana side (maxUpdateRetries) * 3 on es side (default in so client) = 9. Since we removed the custom retry code on the Kibana side, to stick to older behavior it makes sense to make it 9 on es side. Rounding to 10 for a nicer number.
4. Improve batching/debounce session saved object updates code:
   4.1. Group batching/debounce per sessionId
   4.2. Make batching explicit for trackId call only because this is where we actually need this optimization. (After the user saves the session all search requests are immediately polled to be saved into the just created search-session saved object)
5. Some debug logs improvements