Add workflow for debugging slow queries to AI skill

We already have a (out of date) doc internally for this. We need to take this approach and essentially modernize it for e.g. create perfetto table and then make it a full e2e workflow.

The issue names no repository file or test. Start by finding the AI skill and the outdated slow-query workflow referenced in the description, then establish the current approach for creating a Perfetto table. Done means the skill documents a modern, complete end-to-end workflow for debugging slow queries.

Written by the indexing model from the issue text.