tests: add workaround for regression in 1.11

[aplavin]

Good news: 1.11 doesn't really break anything in Accessors, not even any @inferred test :)

The only (surprisingly, still unresolved) issue affecting us is JuliaLang/julia#54664. There's already a PR #170 to work it around, but that one uses internals. I asked on slack today and got a recommendation to just load REPL – it works indeed!

[aplavin]

have no idea about the "invalidations" CI test failing...

[aplavin]

I think @ranocha added invalidations test in #73. Any ideas on what's going on with it and how to fix?

[jw3126]

My hope would be that it gets fixed upsteam, so I could live with merging and this test failing for a while. If it never gets fixed upsteam and no brave soul wants to look into it, we can remove the test.
The fancy tooling is here to serve us, not the other way round.

[jw3126]

you want to bump the patch version?

[ranocha]

Yeah, it looks like it's an upstream error in JET.jl: https://github.com/JuliaObjects/Accessors.jl/actions/runs/11242478221/job/31256238698?pr=175#step:5:449
Did you report it there? From my point of view, going ahead with this PR even when the upstream error persists is fine.

[aplavin]

Ok let's just ignore invalidations CI for now.

you want to bump the patch version?

Not sure, why? It's just a test fix, seems like no need to update the new version for every user when nothing changed in the package...