Extended metrics for sorts

### Is your feature request related to a problem or challenge?

During benchmarking I've realized that the sort operations(especially the complex ones like multi-level sorts) don't really give me a lot of information to go on(generally spill info and the baseline metrics)
I wanted to look into what happens in the actual sort and what takes time there, so I created additional metrics which I found helpful and I think can be useful for anyone else (I'll submit a PR shortly).

### Describe the solution you'd like

What I found useful were timers regarding the evaluation of the actual sort columns, the actual sort on the indices, and the following take algorithm.
but this can be extended further, into "how many levels were in this multi-level merge", "how much time were we awaiting spill tasks", or any other metrics that comes to mind.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_

No file, test, or entry point is named. Start by locating the sort implementation and its existing spill and baseline metrics, then trace sort-column evaluation, index sorting, the take algorithm, and spill-task waits. Done means the selected sort timings or counters are exposed consistently and covered by relevant tests or benchmarking.

Written by the indexing model from the issue text.