[kernel] Save ring3 rflags in cpu_data, not just stack

So that kernel code can modify user rflags, save it in the CPU state
data, and save that off to the TCB when switching tasks.

src/kernel/cpu.h

@@ -28,6 +28,7 @@ struct cpu_data
     uint32_t reserved;
     uintptr_t rsp0;
     uintptr_t rsp3;
+    uint64_t rflags3;
     TCB *tcb;
     thread *thread;
     process *process;