core: optimize convert arrays

[kennykos]

Description

convert_arrays was taking ~2.2 ms for 6 CuPy arrays, versus ~0.1 ms of actual time-in-workunit.

This PR applies the optimizations enumerated in the commit history to bring convert_arrays down to ~0.7 ms, a 3x speedup.

Remaining bottleneck: _init_view in pykokkos/interface/views.py now dominates the remaining cost at ~0.5 ms per array. I think that closing that gap will require optimizing Kokkos view initialization itself, which is out of scope here and left for a follow-up PR.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactor / code cleanup
• Other: runtime optimization

Checklist

• Code is formatted with black==25.12.0 (CI will enforce this)
• Tests pass locally with python3 runtests.py
• New or updated unit tests added in the test directory where applicable
• Examples added in examples/pykokkos for large feature changes
• PR title follows the module: feature naming convention

Testing

Tested with standard test suite.

Additional Notes

Function and kernel timings where done with the tau scripts

#!/bin/bash

echo "Uninstrumented: "

python3 script.py

echo "With TAU:"
export TAU_TRACE=1
/bin/rm -f *.trc *.edf 

tau_python script.py

tau_treemerge.pl

tau_trace2json tau.trc tau.edf -chrome -ignoreatomic -o "convert_arrays_$(date +"%Y%m%d").json"

#firefox https://perfetto.dev &