Way to set compiler optimization flags for native libraries (e.g. -march, -mtune, etc)?

[qwertychouskie]

NativeAOT builds include separate libHarfBuzzSharp.so and libSkiaSharp.so libraries; when targeting a slow device (e.g. Raspberry Pi 3) it'd be nice to compile these with options tuned to the architecture and needs of the project, e.g. something liek this for an rpi3: -mcpu=cortex-a53 -mtune=cortex-a53 -Os -ftree-vectorize -ffast-math -flto -mno-track-speculation -mharden-sls=none

Also, if there are other native libs statically included in the binary, it'd be good to compile them with the same flags.