arm_interface: Add missing fpsr/tpidr members to the ThreadContext struct

Internally within the kernel, it also includes a member variable for the
floating-point status register, and TPIDR, so we should do the same here to match
it.

While we're at it, also fix up the size of the struct and add a static
assertion to ensure it always stays the correct size.

src/core/arm/arm_interface.h

@@ -22,10 +22,16 @@ public:
         std::array<u64, 31> cpu_registers;
         u64 sp;
         u64 pc;
-        u64 pstate;
+        u32 pstate;
+        std::array<u8, 4> padding;
         std::array<u128, 32> vector_registers;
-        u64 fpcr;
+        u32 fpcr;
+        u32 fpsr;
+        u64 tpidr;
     };
+    // Internally within the kernel, it expects the AArch64 version of the
+    // thread context to be 800 bytes in size.
+    static_assert(sizeof(ThreadContext) == 0x320);
 
     /// Runs the CPU until an event happens
     virtual void Run() = 0;

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -247,15 +247,19 @@ void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
     ctx.pstate = jit->GetPstate();
     ctx.vector_registers = jit->GetVectors();
     ctx.fpcr = jit->GetFpcr();
+    ctx.fpsr = jit->GetFpsr();
+    ctx.tpidr = cb->tpidr_el0;
 }
 
 void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
     jit->SetRegisters(ctx.cpu_registers);
     jit->SetSP(ctx.sp);
     jit->SetPC(ctx.pc);
-    jit->SetPstate(static_cast<u32>(ctx.pstate));
+    jit->SetPstate(ctx.pstate);
     jit->SetVectors(ctx.vector_registers);
-    jit->SetFpcr(static_cast<u32>(ctx.fpcr));
+    jit->SetFpcr(ctx.fpcr);
+    jit->SetFpsr(ctx.fpsr);
+    SetTPIDR_EL0(ctx.tpidr);
 }
 
 void ARM_Dynarmic::PrepareReschedule() {

src/core/gdbstub/gdbstub.cpp

@@ -250,7 +250,7 @@ static void RegWrite(std::size_t id, u64 val, Kernel::Thread* thread = nullptr)
     } else if (id == PC_REGISTER) {
         thread->context.pc = val;
     } else if (id == PSTATE_REGISTER) {
-        thread->context.pstate = val;
+        thread->context.pstate = static_cast<u32>(val);
     } else if (id > PSTATE_REGISTER && id < FPCR_REGISTER) {
         thread->context.vector_registers[id - (PSTATE_REGISTER + 1)][0] = val;
     }