Prevent flaky roundtrips in rerun rrd compare

[teh-cmc]

Store-side compaction has election heuristics that can yield different results depending on the order in which the chunks arrive.
This can make some cross-language roundtrip tests flaky.

This PR makes sure to disabled store-side compaction when running RRD comparison tests.

• DNM: requires Automatically generate partial update APIs for eager archetypes #8647

[github-actions]

Web viewer failed to build.

| Result | Commit | Link | Manifest |
| ------ | ------- | ----- |
| ❌ | | https://rerun.io/viewer/pr/8648 | +nightly +main |

^{Note: This comment is updated whenever you push a commit.}

[Wumpf]

I'm rather surprised that compaction isn't deterministic. Is the problem that two rrd files with differing message order should still yield the same store from the pov of a comparision, but the fact that data comes in different order changes compaction?

[teh-cmc]

I'm rather surprised that compaction isn't deterministic. Is the problem that two rrd files with differing message order should still yield the same store from the pov of a comparision, but the fact that data comes in different order changes compaction?

Store-side compaction is currently done synchronously on the write path: when a Chunk arrives, we find the best possible candidate to merge that Chunk into. The result of this process can vary depending on the order the chunks arrive in.