chore: slow test refactoring findings

[dvdplm]

Investigating a slow test I found a few things, collected in this PR.

Unclear if we want to merge this as is. Four classes of changes:

1. Attempt to dial down the logging noise a bit. Useful for debugging but we should might want to do a full sweep instead. Commits
2. Add some tracing based tooling to help in debugging slow tests. Commit.
3. Refactor the CRS testing tooling a bit for speed and readability. Commit.
4. Instrument the test in question with the above new tooling and make it go fast. Commit.

Based on #451 (which in turn is based on #440, which is based on #434), hence the draft status.

So, for reviewers:

• Do we want a full, critical look at our log levels? Root out println!s, demote many INFOs to DEBUG or TRACE etc?
• Are tools like the TimingLayer generally useful or can we do better? Or is it not useful enough?
• Should I continue looking at slow tests and see if we can spot a slowness-inducing pattern and fix them all?
• Is this PR mergeable as-is (once the downstream PRs are merged), or should I split it up?