It's possible for there to be multiple RSDP tables in memory, but only a single one of them should have a valid checksum. In addition to the RSDP, this also adds checksum validation for the rest of the tables (which is now done as a part of the signature check). This doesn't seem to lead to any behavior change when running in QEMU, but might have an impact on real hardware or other virtualizers.

I've also included an extra commit that I had around locally that just asserts that s_hardware_timers is empty before initializing the PIT and RTC since we move on to assume that that was in fact the case.