Kernel: store/load sse/mmx/fpu state on isr/irq/syscall

I haven't tested this yet but should be fine. This will be optimized
to only save state from threads that are using it

kernel/include/kernel/Thread.h

@@ -82,6 +82,9 @@ namespace Kernel
 
 		bool is_userspace() const { return m_is_userspace; }
 
+		void save_sse() { asm volatile("fxsave %0" :: "m"(m_sse_storage)); }
+		void load_sse() { asm volatile("fxrstor %0" :: "m"(m_sse_storage)); }
+
 	private:
 		Thread(pid_t tid, Process*);
 		void on_exit();
@@ -111,6 +114,8 @@ namespace Kernel
 
 		uint64_t m_terminate_blockers { 0 };
 
+		alignas(16) uint8_t m_sse_storage[512];
+
 		friend class TerminateBlocker;
 		friend class Scheduler;
 	};