I'd consider a VM, and I've used them before, but I've found Docker gives the advantage of forcing me to be thorough in my toolchain docs, and portable to CI pipelines.

chroot sounds like nothing but a headache.

QEMU VM, I don't even understand the potential advantage of lol

QEMU for emulating different hardware I guess. Actually I was wondering how you test firmware repeatably without hardware emu; never dealt in firmware myself.

The only way you really can is with hardware in the loop, in my opinion.

People try with mocks and stubs, but hardware in the loop is the only way you can know your firmware works as expected on device.