I suspect you’d need a privileged container for that.
LXD containers have all capabilities enabled out of the box, but those are restricted by the kernel to what’s owned by the user namespace inside of unprivileged containers.
Docker containers are effectively always privileged in that regard, they just have some capabilities or other permissions stripped away.
I don’t believe there is. This may be something that could be eventually done through system call interception but that may get unpleasant very quickly as syscall interception isn’t particularly fast so getting that in the memory access path would be problematic.