Signed 16bit stream incorrectly converted

[xeoncross]

Reading a 2-channel, signed 16bit PCM stream doesn't work (common) because the bytes are converted to int32 instead of int16

n = int32(s[0]) + int32(s[1])<<8

n = int16(buf[0]) + int16(buf[1])<<8

You can see a copy of the byte stream graphed here: http://imgur.com/a/TVuH6

[xeoncross]

We might want to change the logic for 1 or 2 byte sequences to calculate as a signed 16bit number and then type cast to 32 bit.

[xeoncross]

Consider this conversion to float32 using int64 for the full range of values from more than WAV audio sources.

[cryptix]

Looks like @nathan-cormier also ran into this? (re cryptix/wav#14)

[cryptix]

@Xeoncross the link you posted was based on a branch and now points to some unrelated code.. :'( can you update it?