default velocity-to-attenuation curve is incorrect

[mrbumpy409]

The default velocity-to-attenuation curve for a SoundFont voice should be 96 dB deep. TinySoundFont, however, uses a 48 dB curve, reducing the velocity-controlled dynamic range by half.

Here is a screenshot of the recording of Test 13-A in the SoundFont Spec Implementation Test (TinySoundFont on top, FluidSynth on bottom):

You can see that TinySoundFont is not reducing the sample volume enough as the note-on-velocity decreases.

[firodj]

wow amazing works for creating the spec benchmarks.
As of spec24, the velocity-to-attenuation is using modulator with negative concave curve fashion, but currently tsf lacks of the curve and modulators.

I've also implement the tsf modulator on my forks. its using default modulators to do velocity-to-attenuation.