fix(supraseal): add Turing (sm_75) back to CUDA architectures (#1038)

When CUDA 13 support was added, Volta (sm_70) was correctly removed as
CUDA 13 dropped it. However, Turing (sm_75) was also removed even though
CUDA 13 still supports it.

This causes a segfault on `./curio --version` when built on machines with
Turing GPUs (GTX 1650/1660, RTX 2060-2080, Tesla T4). The CUDA kernels
in pc2.cu are compiled only for sm_80+, and the CUDA runtime crashes
during binary initialization when no compatible kernel is found for sm_75.

Binaries built on Ampere+ machines work fine because the embedded kernels
match. But building locally on a Turing machine produces a broken binary.

Adds -gencode arch=compute_75,code=sm_75 back to CUDA_ARCH.

Author: Reiers

extern/supraseal/build.sh

@@ -179,9 +179,11 @@ export PATH=$CUDA/bin:$PATH
 
 echo "Found CUDA $CUDA_VERSION at: $CUDA"
 SPDK="deps/spdk-v24.05"
-# CUDA 13 architectures - removed compute_70 (Volta) as it's no longer supported in CUDA 13+
+# CUDA architectures — include all supported generations.
+# CUDA 13 dropped Volta (sm_70) but still supports Turing (sm_75).
+# sm_75: Turing (GTX 1650/1660, RTX 2060-2080, T4)
 # sm_80: Ampere (A100), sm_86: Ampere (RTX 30xx), sm_89: Ada Lovelace (RTX 40xx, L40), sm_90: Hopper (H100)
-CUDA_ARCH="-arch=sm_80 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_89,code=sm_89 -gencode arch=compute_90,code=sm_90 -t0"
+CUDA_ARCH="-arch=sm_75 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_89,code=sm_89 -gencode arch=compute_90,code=sm_90 -t0"
 CXXSTD=`$CXX -dM -E -x c++ /dev/null | \
         awk '{ if($2=="__cplusplus" && $3<"2017") print "-std=c++17"; }'`