LibGfx: Make FastBoxBlurFilter go faster

When entering FastBoxBlurFilter::apply_single_pass(), we VERIFY that the
bitmap format is BGRA8888.

This invariant allows us to use get_pixel<BGRA8888>() instead of the
generic get_pixel() throughout the function. This removes a lot of
branches and makes the filter significantly faster. :^)

Userland/Libraries/LibGfx/Filters/FastBoxBlurFilter.h

@@ -41,7 +41,7 @@ public:
             sum_red = sum_green = sum_blue = sum_alpha = 0;
             // Setup sliding window
             for (int i = -radius; i <= radius; i++) {
-                auto color_at_px = m_bitmap.get_pixel(clamp(i, 0, width - 1), y);
+                auto color_at_px = m_bitmap.get_pixel<StorageFormat::BGRA8888>(clamp(i, 0, width - 1), y);
                 sum_red += red_value(color_at_px);
                 sum_green += green_value(color_at_px);
                 sum_blue += blue_value(color_at_px);
@@ -57,8 +57,8 @@ public:
                 auto leftmost_x_coord = max(x - radius, 0);
                 auto rightmost_x_coord = min(x + radius + 1, width - 1);
 
-                auto leftmost_x_color = m_bitmap.get_pixel(leftmost_x_coord, y);
-                auto rightmost_x_color = m_bitmap.get_pixel(rightmost_x_coord, y);
+                auto leftmost_x_color = m_bitmap.get_pixel<StorageFormat::BGRA8888>(leftmost_x_coord, y);
+                auto rightmost_x_color = m_bitmap.get_pixel<StorageFormat::BGRA8888>(rightmost_x_coord, y);
 
                 sum_red -= red_value(leftmost_x_color);
                 sum_red += red_value(rightmost_x_color);
@@ -90,7 +90,7 @@ public:
                     sum_blue / div,
                     sum_alpha / div);
 
-                m_bitmap.set_pixel(x, y, color);
+                m_bitmap.set_pixel<StorageFormat::BGRA8888>(x, y, color);
 
                 auto topmost_y_coord = max(y - radius, 0);
                 auto bottommost_y_coord = min(y + radius + 1, height - 1);