GPU: Implement skip-drawing-to-active-field for interlaced mode

Currently only in the software renderer.

src/core/gpu.cpp

@@ -240,7 +240,7 @@ u32 GPU::ReadRegister(u32 offset)
       // code can be dependent on the odd/even bit, so update the GPU state when reading.
       // we can mitigate this slightly by only updating when the raster is actually hitting a new line
       if (IsRasterScanlinePending())
-        m_tick_event->InvokeEarly(true);
+        Synchronize();
 
       return m_GPUSTAT.bits;
     }

src/core/gpu.h

@@ -400,6 +400,12 @@ protected:
       static constexpr u32 MASK = (1 << 19) | (1 << 22);
       return ((bits & MASK) == MASK);
     }
+    bool SkipDrawingToActiveField() const
+    {
+      static constexpr u32 MASK = (1 << 19) | (1 << 22) | (1 << 10);
+      static constexpr u32 ACTIVE = (1 << 19) | (1 << 22);
+      return ((bits & MASK) == ACTIVE);
+    }
 
     // During transfer/render operations, if ((dst_pixel & mask_and) == 0) { pixel = src_pixel | mask_or }
     u16 GetMaskAND() const

src/core/gpu_commands.cpp

@@ -306,6 +306,9 @@ bool GPU::HandleRenderCommand(const u32*& command_ptr, u32 command_size)
                   primitive_names[static_cast<u8>(rc.primitive.GetValue())], ZeroExtend32(num_vertices),
                   ZeroExtend32(words_per_vertex));
 
+  if (m_GPUSTAT.SkipDrawingToActiveField() && IsRasterScanlinePending())
+    Synchronize();
+
   DispatchRenderCommand(rc, num_vertices, command_ptr);
   command_ptr += total_words;
   m_stats.num_vertices += num_vertices;

src/core/gpu_sw.cpp

@@ -574,6 +574,9 @@ void GPU_SW::ShadePixel(u32 x, u32 y, u8 color_r, u8 color_g, u8 color_b, u8 tex
   if ((bg_color.bits & mask_and) != 0)
     return;
 
+  if (m_GPUSTAT.SkipDrawingToActiveField() && BoolToUInt32(m_GPUSTAT.drawing_even_line) != (static_cast<u32>(y) & 1u))
+    return;
+
   SetPixel(static_cast<u32>(x), static_cast<u32>(y), color.bits | m_GPUSTAT.GetMaskOR());
 }