GCC: do not initialize NEON int32x4_t with braces initializer

GCC does not support int32x4_t loading using braces wrapping 4 ints.
Instead we should use the NEON intrinsic to load from an array of
ints.

Bug: 819294
Change-Id: I13b877405273e4ebcc944d50c155ee29ff31cc99
Reviewed-on: https://chromium-review.googlesource.com/951773
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Cr-Commit-Position: refs/heads/master@{#541544}

skia/ext/convolver_neon.cc

@@ -23,7 +23,7 @@ AccumRemainder(const unsigned char* pixels_left,
     remainder[2] += coeff * pixels_left[i * 4 + 2];
     remainder[3] += coeff * pixels_left[i * 4 + 3];
   }
-  return {remainder[0], remainder[1], remainder[2], remainder[3]};
+  return vld1q_s32(remainder);
 }
 
 // Convolves horizontally along a single row. The row data is given in
@@ -336,4 +336,4 @@ void ConvolveVertically_Neon(const ConvolutionFilter1D::Fixed* filter_values,
   }
 }
 
-}  // namespace skia
+}  // namespace skia
\ No newline at end of file