This PR takes inspiration from #306 to update the build process of the binaries in the llvm image:

main difference is that the resulting binaries are statically compiled. This allows them to run in a scratch image, drastically reducing the size of the docker image

The core differences are:

• -DLLVM_BUILD_STATIC="ON" -> ensures the produced binaries are statically linked
• -DCMAKE_CXX_FLAGS="-s -flto" -> ensures the produced binaries are stripped and use link time optimization

Since the binaries are now all static, they can be delivered in a FROM scratch image instead of the current ubuntu based image. This results in an overall smaller image.

Updating the build flags actually also leads to ~20% smaller binaries, contributing to a smaller final image:

• Current image:

linux/arm64	linux/amd64
/usr/local/bin: 161 Mb	/usr/local/bin: 175 Mb

• With static build:

linux/arm64	linux/amd64
/usr/local/bin: 151 Mb	/usr/local/bin: 168 Mb

• With static build and flto:

linux/arm64	linux/amd64
/usr/local/bin: 126 Mb	/usr/local/bin: 137 Mb

Combined with the switch from an ubuntu base to a scratch one, this reduces the llvm image size by about 50% (from 262Mb to 126Mb for linux/arm64).

Additional changes:

• I updated and extended the CST tests so that every binary in the image has at least:
  • a which test
  • an ldd test
  • a <bin> --version smoke test
• I had to bump the timeout for the image build job as the LLVM image build was taking just a little over the default timeout, especially when using the optimizing flag -flto and building a crossed-compiled arm64 image from the amd64 CI runner.