riscv spawn: halt/reboot must HART_STOP every hart, never SRST

[congwang-mk]

Part of the RISC-V port. Prerequisite for the first spawn boot that reaches userspace.

machine_halt(), machine_power_off() and machine_restart() in arch/riscv/kernel/reset.c end in sbi_srst_reset() (SRST extension) or legacy SBI_EXT_0_1_SHUTDOWN, both of which act on the whole machine. A spawn kernel running the stock path powers off or resets the host.

• Spawn-side override (mirror of mk_spawn_reboot_init() in arch/x86/multikernel/spawn.c:1732): halt/power-off/restart = every hart calls HART_STOP; the last hart too. Never SRST, never legacy shutdown, never emergency_restart() through SRST.
• Skip register_platform_power_off(sbi_shutdown) (sbi.c:151) and the sbi_srst_reboot_nb notifier when booted as a spawn.
• smp_send_stop() parks other harts in wait_for_interrupt() inside the image (arch/riscv/kernel/smp.c:86), the "hart still executing the image we are about to overwrite" hazard. Route the stop handler to HART_STOP in a spawn.
• panic() paths: panic_timeout reboot must also go through the override.
• mk_instance_settle_halted() completion = HART_STATUS == STOPPED for every instance hart.

Test: poweroff and reboot inside a spawn leave the host running and every instance hart STOPPED.

[pro-utkarshM]

i’m taking this as part of the #23-#26 RISC-V lifecycle and loader series. i’ll coordinate with @nickolaev and stack the work on #22 once the architecture foundation is available.