chore: CI workaround for pixi 0.68

[aldbr]

See if it would solve #8533

BEGINRELEASENOTES
*CI
FIX: CI workaround for pixi 0.68
ENDRELEASENOTES

[aldbr]

Solving sphinx issues by adding them to no_inherited_members, but opened psf/requests#7459 to fix them upstream

[aldbr]

Clearly not the fix but a workaround (CI can pass).
I couldn't easily freeze pixi 0.67 here because the diracx pixi lock file is already v7.
Using pixi add commands affect gubbins somehow and make pixi panick (to investigate with them):

thread 'main2' (2481) panicked at crates/pixi_core/src/lock_file/resolve/build_dispatch.rs:452:17:
could not initialize build dispatch correctly
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Error:   × failed to solve the pypi requirements of environment 'gubbins' for platform 'linux-64'
  ╰─▶ build dispatch initialization failed: failed to link python-3.11.15-hd63d673_0_cpython.conda

[fstagni]

Do you think to merge this now, and fix properly later? At least we can go ahead with the other PRs.

[aldbr]

Do you think to merge this now, and fix properly later? At least we can go ahead with the other PRs.

Yes this is the idea indeed

[chaen]

@chrisburr was right to insist, pixi.lock solves all our problems and makes it very reproducible: it fails everywhere for everybody and we can't release, so no new bug in productions !! 🧌 🤣