Restore PCM sample conversion.

As shipped, projectM converts incoming integer PCM samples to 32-bit
floats in the range [-2.0, 2.0].  Thinking this was an error, I fixed
this to convert to the range [-1.0, 1.0] (because signal processing
routines "obviously" expect their inputs to be in this range so they can
play all kinds of games with unit vectors, right?)  Whether or not the
range [-1.0, 1.0] is actually preferable, unfortunately all the shipping
presets were written to expect the range [-2.0, 2.0], with the
consequence that they look less interesting if they operate on the
reduced range.

Change the sample conversion back such that 16-bit integer samples are
once again converted to floats in the range [-2.0, 2.0].

Change-Id: I89dbb6a2a758ca3c3874029ef8eb34ceed32fde2
1 file changed