CLUE implementation in GPU gives only outliers

[ebrondol]

After the merging of #35, the implementation on GPU will return no clusters, but only outliers.
This currently does not break the tests (which check only if the code run without errors), but it needs to be fixed with a subsequent PR.

To test it:

# source key4hep environment
source /cvmfs/sw.hsf.org/key4hep/setup.sh

# get nvcc 11.4, if needed
source /cvmfs/sft.cern.ch/lcg/contrib/cuda/11.4/x86_64-centos8/setup.sh

# then setup this project
git clone --recurse-submodules https://github.com/key4hep/k4Clue.git
cd k4Clue
cmake -S . -B build
cmake --build build

./build/src/clue/main data/input/aniso_1000.csv 20 25 2 1 1 1

[ebrondol]

As additional information, I have tried to add the following printouts: https://github.com/ebrondol/k4Clue/compare/geometryTemplateGPU_rebase...ebrondol:k4Clue:GPUdebug?expand=1
which they work fine before the PR, but after the template modifications, the ones in the kernel_calculate_density functions are not printed out.