sw_engine: small optimization with a fast-clipping region access

Now the engine can access intermidate rle sections
with a specific y ranges, It doesn't need full iterations.

src/renderer/sw_engine/tvgSwRle.cpp

@@ -881,12 +881,18 @@ bool rleClip(SwRle *rle, const SwRle *clip)
     Array<SwSpan> out;
     out.reserve(std::max(rle->spans.count, clip->spans.count));
 
-    auto spans = rle->data();
+    const SwSpan *end;
-    auto end = rle->spans.end();
+    auto spans = rle->fetch(clip->spans.first().y, clip->spans.last().y, &end);
-    auto cspans = clip->data();
-    auto cend = clip->spans.end();
 
-    while(spans < end && cspans < cend) {
+    if (spans >= end) {
+        rle->spans.clear();
+        return false;
+    }
+
+    const SwSpan *cend;
+    auto cspans = clip->fetch(spans->y, (end - 1)->y, &cend);
+
+    while (spans < end && cspans < cend) {
         //align y-coordinates.
         if (cspans->y > spans->y) {
             ++spans;
@@ -928,9 +934,10 @@ bool rleClip(SwRle *rle, const RenderRegion* clip)
     Array<SwSpan> out;
     out.reserve(rle->spans.count);
     auto data = out.data;
+    const SwSpan* end;
     uint16_t x, len;
 
-    ARRAY_FOREACH(p, rle->spans) {
+    for (auto p = rle->fetch(*clip, &end); p < end; ++p) {
         if (p->y >= max.y) break;
         if (p->y < min.y || p->x >= max.x || (p->x + p->len) <= min.x) continue;
         if (p->x < min.x) {