Remove some computations in histogram clustering

- move the bin_id to the Histogram
- do not consider empty histograms

The speed-ups are negligible as linear algorithms in uint16_t are
removed, while the whole code is still O(N^2) in histograms.

Change-Id: Ie9c4831f0f3c64af9d9710a1dc2d817ba165389e

src/enc/histogram_enc.h

@@ -44,6 +44,7 @@ typedef struct {
   uint64_t red_cost;        // literal, red & blue.
   uint64_t blue_cost;
   uint8_t is_used[5];       // 5 for literal, red, blue, alpha, distance
+  uint16_t bin_id;          // entropy bin index.
 } VP8LHistogram;
 
 // Collection of histograms with fixed capacity, allocated as one