monitoring: use heatmap for aggregate successful operation durations

Created by: Strum355

This work is adjacent to https://github.com/sourcegraph/sourcegraph/pull/27171, and introduces what I believe to be more useful data than the aggregate 99th percentile. We still retain the grouped 99th percentile immediately below, while introducing a duration distribution view.

For cases where the buckets are incorrectly configured (as per the top panel), the results look essentially the same.

Before vs After: