torch_cluster: fix for compiling with py 3.11 and cuda 12.4
To run the testing, add a comment with:
Bot test PACKAGE[ PACKAGE2 ...]|all in TOOLCHAIN|all [(only|also) for ARCH-LABEL-COMPILER-BUILDTYPE] please!
later followed by
Bot repeat please!
To ensure patches required for existing LCG branches are not lost, please add the respective labels: alsoTargeting:LCG_102-patches, alsoTargeting:master, etc.