Migrate internal duration/performance logging

[janpio]

Issues like #13692 or #12384 are currently hard to understand and debug because we have no debug logging about how long the individual sub steps of Migrations CLI command take. We should add that, so we can ask users reporting such problems to enable it and report to us.

The goal is to already have a better understanding of what exactly is slow (is it one specific step? or are just all steps generally slow?) before going deeper into reproduction or fixing.

[tomhoule]

The migration engine already has a lot of instrumentation we use for debugging in tests. We've had for a long time tracing spans and events using the same tracing framework the query engine recently adopted. It could be as simple as figuring out a non-intrusive, nice way to expose timings for these spans. That would include how long each rpc request takes, as well as more granular spans inside each command.