This PR fixes the flakiness of our replay event flushing test by slightly changing the test requirements/expectations:

• Disabled compression to avoid potential timing problems
• Instead of asserting that a flush happens within a certain time frame from the last flush, we ease up this expectation so that we only expect that another flush happens after the previous one, given that click events happened in between.
• Instead of forcing a flush by triggering a visibility change (like we do a lot in other tests to avoid flakes), this test focuses on the max delay of flushing, thereby ensuring that our flushes are not debounced to eternity.

Note: I tested that flakes were gone by running this test 75x per CI job for multiple times and didn't see this test failing anymore.

closes #7266