kafka replay speed: fix concurrent fetching concurrency transition

[dimitarvdimitrov]

What this PR does

fixes

1. There's a race between Add()ing to the WaitGroup and then Wait()ing for it; it's possible that we close(r.done) and Wait(); then Wait() just returns immediately
2. If we immediately call Update() after newConcurrentFetchers, then lastReturnedRecord might stay as -1. So we'd resume from offset -1 🥴 instead of from the start or end of the partition
3. after Update() we should continue from lastReturnedRecord+1 because we've already fetched and returned lastReturnedRecord; this isn't critical though

This is missing tests; WIP

Which issue(s) this PR fixes or relates to

Fixes #

Checklist

• [ ] Tests updated.
• [ ] Documentation added.
• [ ] CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX].
• [ ] about-versioning.md updated with experimental features.