[TDX] Improve TLB lock code

[smalis-msft]

• Remove the unneeded heap allocation and double copy
• add additional checks to prevent unneeded waking
• verify that the extended check that's currently missing isn't actually needed (it was)
• Replace the RwLock with a purely atomic ring buffer (only needed for perf, not functionality)
• Add hcl_cancel_mask ioctl and use it for waking target_vps (so we can properly wait for them to enter VTL 2)
• Clear VP entry flags correctly
• Lazy processor set parsing
• Reenable CPUID (with_use_hypercall_for_remote_flush_and_local_flush_entire)
• Consider tweaking CPUID on revoke
• Perf test and see if parts need to move into the kernel