Ability to query the audio volume currently coming out of the synth?

[derselbst]

During discussion of #1159 it turned out that gapless playback can not be easily achieved because the articulation of reverb may be too dynamic. Instead, an idea was made to add the ability to query the audio volume coming out of the synth. There are several questions related to this:

1. Audio volume or measuring loudness in general is an open field with many different approaches on how to do that. Should such functionality really be integrated in fluidsynth? (Client applications could easily measure volume by themselves by just inspecting the audio buffer rendered by fluidsynth.)
2. If such a "convenience functionality" should be integrated into fluidsynth, which one to choose? (RMS?, Sample Peak?, EBUR128?)
3. How should the audio channels be treated, if multi-channel rendering is enabled? I.e. derive a volume for each channel separately or mixing it all down to get a single value?

Note that this "query-audio-volume" function should not rely on any data currently cached in fluidsynth. That's mainly due to race conditions when querying this function from outside the synth context, while concurrently the synth thread might run making changes to the audio buffers. And secondly, because ideally I wanted to get rid of internal caching at all and directly render onto user-provided buffers.