#2322 Strata v0.1.44 fails to compile on Linux with CUDA Toolkit 12.1.

Description

Strata v0.1.44 fails to compile on Linux with CUDA Toolkit 12.1.

Environment:
- Strata tag: v0.1.44
- Commit: ed4773eb
- GPU: RTX 3090 + RTX 3060
- CUDA architecture: sm_86
- NVIDIA driver: 595.99.02
- nvcc: CUDA 12.1, V12.1.105
- CMake/Ninja build through update.sh

Fatal errors:
src/qwen35/q36_engine.cu:809:
calling a __device__ function "__ushort_as_half(unsigned short)"
from a __host__ function "upload_f32" is not allowed

src/qwen35/q36_engine.cu:814:
same error in upload_f32

The updater retries and correctly retains engine 0.1.43.1.
CUDA 12.1 is detected as supported by setup, but the new
strata_qwen35 target does not compile.



"hercules@Hercules:~/Strata/Strata_Linux$ nvcc --version
git rev-parse HEAD
gcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Mon_Apr__3_17:16:06_PDT_2023
Cuda compilation tools, release 12.1, V12.1.105
Build cuda_12.1.r12.1/compiler.32688072_0
ed4773eb7c8f1f1531ba2bc70e0e1987de2661af
gcc (Ubuntu 12.4.0-2ubuntu1~24.04.1) 12.4.0
Copyright (C) 2022 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

hercules@Hercules:~/Strata/Strata_Linux$ 
"