Improve performance of fetching bundled aggregations

[clokep]

As detailed in the jaeger traces from #11611 the bundled aggregations code calculates the bundled aggregations for each event to be returned. To calculate the bundled aggregations for a single event is currently requires 6 queries (across 5 transactions).

For some APIs this is (e.g. /events only returns a single event), but /sync or /messages may return many events causing an explosion of queries.

Some of the fetched data is fairly simple and we should be able to fetch the data for multiple events at once, which should save queries. #11660 has some basic timing info that this might be an improvement, I imagine reducing the round-trips from a remote database would cause the gap to be further improved.

[clokep]

I think I've done everything which is reasonable here? Unfortunately fetching the necessary info for m.reference and m.annotation relations looks quite hard to do as bulk queries.

I'm going to call this done since #11660 and #11752 both improved performance quite a bit.