[arm64] Add support for chain_load for arm64 platforms

[mpayam]

Tested on VIM2 by loading LK from LK.

This is phase 1. Will need to add support to go back to initial boot exception level (as opposed to booting the next stage in EL1).

[travisg]

Closing this in favour of #526, which records what's actually left to do.

To be clear about why: this is not superseded. arch_chain_load() is still PANIC_UNIMPLEMENTED on arm64 today, so the gap you were filling is real and open. But the branch is 1550 commits behind and no longer cherry-picks cleanly, and re-reading it against your own note about phase 2, the remaining work is the hard part rather than a rebase.

Specifically, arm_chain_load hardcodes sctlr_el1 with the comment "LK runs in EL1", clears SCTLR.M with an isb but no dsb or TLB invalidate, and leaves SCTLR.C/SCTLR.I alone. Getting back to the boot exception level — the phase 2 you flagged — isn't symmetric with arm64_elX_to_el1 dropping down, and needs either a stashed return path or firmware help. The scratch-aspace trick you used to keep the loader identity mapped across the MMU-off transition is the right idea and I've written it into the issue so it isn't lost.

Thanks for the original work, and sorry it sat unreviewed for this long.