CUDA: add numeric conversion tests\n\nAdds kernel-based tests

[CodersAcademy006]

This PR adds kernel-based tests for device-side numeric conversions in Numba-CUDA:

• int ↔ float

• float/int → complex

• bool ↔ numeric

Tests are skipped under cudasim due to divergent semantics, and NumPy is used as the reference oracle. The scope is intentionally limited to safe, well-defined CUDA conversions (no Python object, string, or shape-changing conversions).

This continues the systematic porting of CPU-side tests to CUDA, directly contributing to issue #515.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Greptile Summary

Adds a new test file for device-side numeric conversions in CUDA kernels. The implementation tests int↔float, float/int→complex, and bool↔numeric conversions using NumPy as the reference oracle.

• Properly includes SPDX license header (addresses previous review comment)
• Uses @skip_on_cudasim decorator to handle divergent semantics
• Clean kernel-based test pattern with helper methods _launch_1d and _run_conversion
• Comprehensive coverage of basic numeric conversion types
• Tests align well with issue Incomplete Test Coverage in Numba-CUDA #515 goals for porting CPU tests to CUDA

Confidence Score: 5/5

• Safe to merge - well-structured test file with proper licensing and clear test coverage
• The PR adds a straightforward test file following established patterns in the codebase. The SPDX header issue from previous review has been addressed. Tests are well-scoped, use appropriate decorators, and follow the repository's testing conventions. No logic issues, security concerns, or anti-patterns detected.
• No files require special attention

Important Files Changed

Filename	Overview
numba_cuda/numba/cuda/tests/test_conversion.py	New test file for CUDA numeric conversions (int↔float, float/int→complex, bool↔numeric). Well-structured with SPDX header, uses NumPy as reference oracle, properly skips cudasim.

[greptile-apps]

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}