cmake.opts contains -march-native switch which conflicts with the building goal of dockerimages.
This might be nice for running optimized.code locally. But the created binary has very narrow application on cpus of the same type or generation or only some generations. But scripts like install.sh, install.py or createdockerimage.sh make direct use cmake.opts, the code inside docker-binary is expected to run on different computers with amd or intel-cpu with much broader range of variation. Think of applications like course environments (you can't know what hardware you get) or computers of random applicants which just want to try out an docker image.
I would suggest for not to have this option inside the default cmake.opts, but as a commented out suggestion for optimization switches. When createdocker.sh scripts and install scripts are overhauled there could be an option for the sake of higher flexibility to pass an individual adapted .opts-file in.