CPU/Recompiler: Emit test/jz instead of bt/jnc for bits < 8

Likely very slightly faster.

src/core/cpu_recompiler_code_generator_x64.cpp

@@ -1990,6 +1990,33 @@ void CodeGenerator::EmitConditionalBranch(Condition condition, bool invert, Labe
 }
 
 void CodeGenerator::EmitBranchIfBitClear(HostReg reg, RegSize size, u8 bit, LabelType* label)
+{
+  if (bit < 8)
+  {
+    // same size, probably faster
+    switch (size)
+    {
+      case RegSize_8:
+        m_emit->test(GetHostReg8(reg), (1u << bit));
+        m_emit->jz(*label);
+        break;
+
+      case RegSize_16:
+        m_emit->test(GetHostReg16(reg), (1u << bit));
+        m_emit->jz(*label);
+        break;
+
+      case RegSize_32:
+        m_emit->test(GetHostReg32(reg), (1u << bit));
+        m_emit->jz(*label);
+        break;
+
+      default:
+        UnreachableCode();
+        break;
+    }
+  }
+  else
   {
     switch (size)
     {
@@ -2013,6 +2040,7 @@ void CodeGenerator::EmitBranchIfBitClear(HostReg reg, RegSize size, u8 bit, Labe
         break;
     }
   }
+}
 
 void CodeGenerator::EmitBindLabel(LabelType* label)
 {