byte
0-3	'vawt'		as big-endian text - all the following bytes are little-endian
4-7	wave_size 	size of each wave; 2 ... 4096, must be a power of 2 (v1.3)
8-9	wave_count 	number of waves; 1 ... 512 (v1.3)
10-11	flags
        0001: is a sample instead of a wavetable
        0002: is a looped sample
        0004: if not set, wave data is in float32 format
              if set, wave data is in int16 format
        0008: if not set, a sample with 0 dBFS peak will end up having a peak of 2^14 (uses 15 bits, -6 dBFS peak)
              if set, int16 data uses the full 16-bit range
12-end	wave data
        float32 format: 4 * wave_size * wave_count bytes
        int16 format:   2 * wave_size * wave_count bytes
