fix precision problem of silent

[pengzhendong]

frame_rate = 16000
duration = 1013
print("Before the pr:", int(frame_rate * (duration / 1000.0)) / frame_rate)
print("After the pr:", int(frame_rate * duration / 1000.0) / frame_rate)

Before the pr: 1.0129375
After the pr: 1.013

If the pr could not be merged, hot fix:

from pydub import AudioSegment
def silent(cls, duration=1000, frame_rate=11025):
    frames = int(frame_rate * duration / 1000.0)
    data = b"\0\0" * frames
    return cls(data, metadata={"channels": 1,
                                "sample_width": 2,
                                "frame_rate": frame_rate,
                                "frame_width": 2})
AudioSegment.silent = classmethod(silent)