fix[cartesian]: Race condition in unit test for K write

[FlorianDeconinck]

The interval analysis in the unit test test_K_offset_write_conditional fails to catch a mistake in the code that leads to a race condition.

Work:

• Fix the bad interval
• Remove not needed restriction on CUDA version

Further work to fix the underlying problem and the larger issue of bound check on variable indexing is covered here and there

[havogt]

I am not any more familiar enough with cartesian. What's the expected behavior? Does the user need to protect the oob in the vertical interval or should gt4py take care but it's buggy?

[FlorianDeconinck]

I am not any more familiar enough with cartesian. What's the expected behavior? Does the user need to protect the oob in the vertical interval or should gt4py take care but it's buggy?

Sorry I put this PR in a rush I should have explained a bit better.

This PR is aimed at stabilizing the CI first and foremost. The issue is dual:

• On vertical access with scalar offset, gt4py should catch the wrong interval and yell. This is buggy and/or not extended to write, I logged it as [cartesian] K-offset writes bound check is failing #1754 for further work
• On vertical access with variable, there's no way we can check, so a system of bound check on debug should be put in place to mitigate (logged as [cartesian]: Mitigation for K offset write #1684)

Now of course this is part of the larger discussion on physics and there's a definite need for us to sit down and probably talk it through as I am very well aware that we are stepping beyond the bounds of the original design of cartesian.

But we are really responding to the need of the algorithm and our current wider strategy is to not rewrite the algorithm