Try to make ingest lock timeout tests less flaky

[zoldar]

Changes

Can't tell for sure as this test fail is hard to reproduce, but I suspect there might be situations where reaching session processing step in the first ingest takes longer than the sleep before spinning up the second ingest.

[zoldar]

@aerosol hmmm, I could track whether there's con_cache lock present with "eventually" in tight loop and only attempt the second ingest after that's satisfied. I'll see how feasible that is.

[aerosol]

doesn't have to be con_cache specifically, since we're preloading the module in test_helper now

[aerosol]

What I did to track it down previously:

1. Run iex in MIX_ENV=test
2. Set up flameon https://github.com/DockYard/flame_on/tree/main - looking at how @ruslandoga did it at https://github.com/ruslandoga/plausible/pull/301
3. Run tests of interests from iex via Mix.Task.run