0.3.0

The most important modifications here are:

1. Implemented SIMD for wavelet and colour transforms, for both the reversible and irreversible paths.
2. Implemented faster trunc and round functions.
3. Fixed a bug in error/warning messaging.

Improvement in speed:

1. Point 1 produced a small improvement in speed.
2. Point 2 produced a decent improvement in speed.
   More speed can be obtained by improving the speed of the block coder, or, for the irreversible path, employing 16-bit arithmetic instead of floats.

Now on an i7-6700:

For irreversible wavelet:
Tested with 4K 4:4:4 image with 8bits depth, using "-qstep 0.007;" this produces a file that has around same size as that produced by JPEG with Q95.
Encoding: 0.27s.
Decoding: 0.22s.

For reversible wavelet:
Same image with lossless compression
Encoding: 0.53s.
Decoding: 0.34s.