Check whether MAXIMUM_ACTION_COUNT_PER_SAVE should depend again on tracing type

[philippotto]

With #7839, the logic for selecting how many actions can be put into one save request got a bit simpler. Previously, the amount of actions was thresholded against constants that depended on the tracing type. E.g., 15000 skeleton actions, but only 3000 volume actions could be put into one save request. This logic got lost, but could get re-introduced. The benefit would be more efficient saves. However, it's not clear whether this is really important.

My suggestion: we should keep an eye on how fast saving performs. Alternatively, we could benchmark it explicitly, but the priority seems rather low. Still, I want to record this topic in this issue.