Option to use speech-optimized tempo effect in from time_stretch()

[babua]

The SoX tempo effect has additional flags -s and -m which optimize WSOLA parameters for speech and music, respectively.

In my testing, I found that using the time_stretch() without the -s flag, there are significant artifacts in the audio, already at 0.7 or 1.3 factors, which are not too extreme, in my opinion. Simply adding the -s flag makes the output much more smoother.

As AudioTools is speech-oriented, I think it would be a sane default for many users to use this flag in time_stretch().

Sample below with original, before, after, code was:

def stretch(wav)
    a = AudioSignal(wav,sample_rate=24000)
    return a.clone().time_stretch(speed,quick=False).audio_data

0.5_0.7_stretch_samples.zip

[pseeth]

Looks like a good change. Do you mind editing the base branch of this PR to point to the one in this PR: #90, so I can merge it in with the next version, and updating the relevant test to cover the change? The test is here: https://github.com/descriptinc/audiotools/blob/master/tests/core/test_effects.py#L170-L181

[babua]

btw, it's been a while but in case it escaped your attention, I'd added the requested changes. if not, no need to reply! :)