multi-thread decoding: ~25-30% faster

To be enabled with the flag WEBP_USE_THREAD.
For now it's only available on unix (pthread), when using Makefile.unix
Will be switched on more generally later.

In-loop filtering and output (=rescaling/yuv->rgb conversion)
is done in parallel to bitstream decoding, lagging 1 row behind.

Example:
examples/dwebp bryce.webp -v
Time to decode picture: 0.680s

examples/dwebp bryce.webp -v -mt
Time to decode picture: 0.515s

Change-Id: Ic30a897423137a3bdace9c4e30465ef758fe53f2

src/webp/decode.h

@@ -351,6 +351,7 @@ typedef struct {
   int scaled_width, scaled_height;    // final resolution
   int force_rotation;                 // forced rotation (to be applied _last_)
   int no_enhancement;                 // if true, discard enhancement layer
+  int use_threads;                    // if true, use multi-threaded decoding
 } WebPDecoderOptions;
 
 // Main object storing the configuration for advanced decoding.

src/webp/decode_vp8.h

@@ -67,10 +67,13 @@ struct VP8Io {
   VP8IoPutHook put;
 
   // called just before starting to decode the blocks.
-  // Should returns 0 in case of error.
+  // Must return false in case of setup error, true otherwise. If false is
+  // returned, teardown() will NOT be called. But if the setup succeeded
+  // and true is returned, then teardown() will always be called afterward.
   VP8IoSetupHook setup;
 
-  // called just after block decoding is finished (or when an error occurred).
+  // Called just after block decoding is finished (or when an error occurred
+  // during put()). Is NOT called if setup() failed.
   VP8IoTeardownHook teardown;
 
   // this is a recommendation for the user-side yuv->rgb converter. This flag