Update a ton of audio code to use size_t more correctly and in general reduce

webrtc/common_audio/resampler/sinc_resampler_sse.cc

Index: webrtc/common_audio/resampler/sinc_resampler_sse.cc
diff --git a/webrtc/common_audio/resampler/sinc_resampler_sse.cc b/webrtc/common_audio/resampler/sinc_resampler_sse.cc
index e00e0e5dbe54b36bed9caacec1fb53b6f8efee5a..9e3953fede9a1f2bb1bde69f43aa6de14ba628e6 100644
--- a/webrtc/common_audio/resampler/sinc_resampler_sse.cc
+++ b/webrtc/common_audio/resampler/sinc_resampler_sse.cc
@@ -27,13 +27,13 @@ float SincResampler::Convolve_SSE(const float* input_ptr, const float* k1,
   // Based on |input_ptr| alignment, we need to use loadu or load.  Unrolling
   // these loops hurt performance in local testing.
   if (reinterpret_cast<uintptr_t>(input_ptr) & 0x0F) {
-    for (int i = 0; i < kKernelSize; i += 4) {
+    for (size_t i = 0; i < kKernelSize; i += 4) {
       m_input = _mm_loadu_ps(input_ptr + i);
       m_sums1 = _mm_add_ps(m_sums1, _mm_mul_ps(m_input, _mm_load_ps(k1 + i)));
       m_sums2 = _mm_add_ps(m_sums2, _mm_mul_ps(m_input, _mm_load_ps(k2 + i)));
     }
   } else {
-    for (int i = 0; i < kKernelSize; i += 4) {
+    for (size_t i = 0; i < kKernelSize; i += 4) {
       m_input = _mm_load_ps(input_ptr + i);
       m_sums1 = _mm_add_ps(m_sums1, _mm_mul_ps(m_input, _mm_load_ps(k1 + i)));
       m_sums2 = _mm_add_ps(m_sums2, _mm_mul_ps(m_input, _mm_load_ps(k2 + i)));