clang 21.1.1+: workaround regression in the `_mm_mask{,z}_fixupimm{,_… #25

Job	Run time
x86-xop	29m 35s
sleef	35m 53s
gcc-16 (ubuntu-24.04-arm)	17m 39s
gcc-qemu (16, riscv64, riscv64, riscv64, ubuntu-24.04)	43m 11s
x86 (-DSIMDE_NATURAL_VECTOR_SIZE=256 -march=x86-64-v3 -mavx512bw -mavx512vl)	26m 21s
gcc-16 (ubuntu-24.04)	28m 13s
x86 (-march=x86-64-v4 -mcx16 -mxsave -mpclmul -mfsgsbase -mrdrnd -mhle -mrdseed -maes -mclflushop...	34m 53s
x86 (-march=x86-64-v4 -mcx16 -mxsave -mpclmul -mfsgsbase -mrdrnd -mhle -mrdseed -maes -mclflushop...	33m 50s
gcc-qemu (16, armv7, arm, eabihf, armhf, ubuntu-24.04)	43m 38s
x86 (-march=x86-64 -maes -mpclmul)	31m 29s
gcc-qemu (16, -no-extras, aarch64, aarch64, arm64, ubuntu-24.04)	38m 23s
gcc-qemu (16, aarch64, aarch64, arm64, ubuntu-24.04)	39m 1s
gcc-qemu (-O3, 16, riscv64, riscv64, riscv64, ubuntu-24.04)	31m 54s
gcc-qemu (16, armel, eabi, armel, arm, ubuntu-24.04)	42m 9s
gcc-16 (ubuntu-24.04, -ffast-math)	25m 21s
x86 (-march=x86-64-v3)	28m 36s
gcc-qemu (16, s390x, s390x, s390x, ubuntu-24.04)	35m 28s
gcc-qemu (16, loongarch64, loongarch64, loong64, ubuntu-24.04)	40m 36s
x86 (-march=x86-64-v2)	27m 0s
x86 (-march=x86-64-v4)	33m 36s
gcc-qemu (-32bit, 16, armv8, arm, eabihf, armhf, ubuntu-24.04)	44m 41s
	11h 51m 27s

Provide feedback