Allocate width aligned to MCU size for boundary rows

jpeg_write_raw_data() expects input size to be aligned to MCU.
If the image dimensions are not aligned then padding is done for
boundary rows. The padded rows allocation is not proper and this
is fixed.

Bug: oss-fuzz-63314
Test: ultrahdr_enc_fuzzer

jpegencoderhelper.cpp

@@ -147,15 +147,15 @@ bool JpegEncoderHelper::compressYuv(jpeg_compress_struct* cinfo, const uint8_t*
     JSAMPROW cr[kCompressBatchSize / 2];
     JSAMPARRAY planes[3]{y, cb, cr};
 
+    const int aligned_width = ALIGNM(cinfo->image_width, kCompressBatchSize);
+    const bool need_padding = (lumaStride < aligned_width);
     size_t u_plane_size = chromaStride * cinfo->image_height / 2;
     uint8_t* y_plane = const_cast<uint8_t*>(yBuffer);
     uint8_t* u_plane = const_cast<uint8_t*>(uvBuffer);
     uint8_t* v_plane = const_cast<uint8_t*>(u_plane + u_plane_size);
-    std::unique_ptr<uint8_t[]> empty = std::make_unique<uint8_t[]>(cinfo->image_width);
-    memset(empty.get(), 0, cinfo->image_width);
+    std::unique_ptr<uint8_t[]> empty = std::make_unique<uint8_t[]>(aligned_width);
+    memset(empty.get(), 0, aligned_width);
 
-    const int aligned_width = ALIGNM(cinfo->image_width, kCompressBatchSize);
-    const bool need_padding = (lumaStride < aligned_width);
     std::unique_ptr<uint8_t[]> buffer_intrm = nullptr;
     uint8_t* y_plane_intrm = nullptr;
     uint8_t* u_plane_intrm = nullptr;
@@ -227,12 +227,12 @@ bool JpegEncoderHelper::compressY(jpeg_compress_struct* cinfo, const uint8_t* yB
     JSAMPROW y[kCompressBatchSize];
     JSAMPARRAY planes[1]{y};
 
-    uint8_t* y_plane = const_cast<uint8_t*>(yBuffer);
-    std::unique_ptr<uint8_t[]> empty = std::make_unique<uint8_t[]>(cinfo->image_width);
-    memset(empty.get(), 0, cinfo->image_width);
-
     const int aligned_width = ALIGNM(cinfo->image_width, kCompressBatchSize);
     const bool need_padding = (lumaStride < aligned_width);
+
+    uint8_t* y_plane = const_cast<uint8_t*>(yBuffer);
+    std::unique_ptr<uint8_t[]> empty = std::make_unique<uint8_t[]>(aligned_width);
+    memset(empty.get(), 0, aligned_width);
     std::unique_ptr<uint8_t[]> buffer_intrm = nullptr;
     uint8_t* y_plane_intrm = nullptr;
     JSAMPROW y_intrm[kCompressBatchSize];