Use cross-compilation to speed-up arm64 build

[kathoef]

Building the arm64 container image takes very long 55 minutes, whereas building the amd64 images takes only about 5 minutes. According to the Docker buildx docs, there are three multi-platform build strategies

1. Using the QEMU emulation support in the kernel
2. Building on multiple native nodes using the same builder instance
3. Using a stage in Dockerfile to cross-compile to different architectures

Currently, QEMU is implemented. Maybe we should leverage a Docker multi-stage build cross-compilation approach, since Github actions don't provide native arm64 nodes?