ci[cartesian]: env var on CSCS-CI to suppress cuda backend deprecation error

[DropD]

Description

Suppress the deprecation error for cartesian "CUDA" backend on CSCS-CI.

The following open questions remain (should be answered for a more long-term solution):

• Why is there a deprecation error, not a warning by default (warning seems to be the convention in GT4Py)?
• Why did CSCS-CI not fail on feat[cartesian]: gtc cuda backend deprecation #1498, only on main after merging?
• Why does only one test fail (test_bad_layout_warns[cuda])? Is the CUDA backend otherwise untested now?

Requirements

• All fixes and/or new features come with corresponding tests.
• Important design decisions have been documented in the approriate ADR inside the docs/development/ADRs/ folder.

[FlorianDeconinck]

I should have seen this issue in my PR, I am not sure why. Apologies.

cuda backend on cartesian not being tested is not an issue, considering we established it wasn't used by anyone when we poked the main users. We keep in there for silent users, if they exists.

The error over warning, comes from the fact that cuda is now going to misbehave on new features, therefore it shouldn't be used.

[DropD]

I should have seen this issue in my PR, I am not sure why. Apologies.

cuda backend on cartesian not being tested is not an issue, considering we established it wasn't used by anyone when we poked the main users. We keep in there for silent users, if they exists.

The error over warning, comes from the fact that cuda is now going to misbehave on new features, therefore it shouldn't be used.

I checked and the CSCS-CI did pass on your PR before merging. I don't understand why at the moment.