devops: docker release switch to using ubuntu-22.04-s390x image

[taronaeo]

The Docker workflow has been failing with the following error message because we are cross-building, and the emulation does not have the required hardware level information from cpuinfo for CMake to detect automatically.

57.59 FAILED: ggml/src/CMakeFiles/ggml-cpu.dir/ggml-cpu/ggml-cpu.c.o 
57.59 ccache /usr/bin/cc -DGGML_BACKEND_BUILD -DGGML_BACKEND_SHARED -DGGML_SCHED_MAX_COPIES=4 -DGGML_SHARED -DGGML_USE_CPU_REPACK -DGGML_USE_LLAMAFILE -DGGML_USE_OPENMP -DGGML_VXE -D_GNU_SOURCE -D_XOPEN_SOURCE=600 -Dggml_cpu_EXPORTS -I/app/ggml/src/.. -I/app/ggml/src/. -I/app/ggml/src/ggml-cpu -I/app/ggml/src/../include -O3 -DNDEBUG -std=gnu11 -fPIC -Wshadow -Wstrict-prototypes -Wpointer-arith -Wmissing-prototypes -Werror=implicit-int -Werror=implicit-function-declaration -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wdouble-promotion -march=native -mtune=native -mvx -mzvector -fopenmp -MD -MT ggml/src/CMakeFiles/ggml-cpu.dir/ggml-cpu/ggml-cpu.c.o -MF ggml/src/CMakeFiles/ggml-cpu.dir/ggml-cpu/ggml-cpu.c.o.d -o ggml/src/CMakeFiles/ggml-cpu.dir/ggml-cpu/ggml-cpu.c.o -c /app/ggml/src/ggml-cpu/ggml-cpu.c
57.59 cc1: error: hardware vector support not available on z900

This PR switches the cross-build to an actual ubuntu-22.04-s390x image, where the hardware level can be detected correctly and built. Also disables the QEMU emulation for cross-build for s390x.