Resolves issue #65 - treat classification byte as unsigned char

[jason-daly]

All 8 bits of the classification byte have the potential to be used, in both the older formats (0..5) and in the newer formats (6..10). The code had been reading this byte into a signed byte, and then casting it to a short.
For the older formats, this would mean that if the 'Withheld' bit was set (bit index 7), this bit would be interpreted as a sign bit, and effectively be lost in the cast (I think!).
For the newer formats, where the entire byte is used for the classification value, any classifications > 127 are being incorrectly read as negative values.

I think the same issue applies in the compressed decoders, because the uncompressed/decoded classification value was being cast down to a signed byte, when it should be treated as an unsigned byte.

Looking at the original C++ copy of this code, a U8 is used for all of the decoding and reading of the classification byte. These code changes are attempting to produce the equivalent unsigned byte treatment in the java.

[mreutegg]

Looks good to me. Thanks for the fix.