fix support for 8-bit audio

This solves the issue where 8-bit audio is inconsistently treated as unsigned and signed integers.
The audioop library only supports signed integers, so this fix converts to and from signed integers where necessary.

It removes places where the values are treated like unsigned as these partial fixes are no longer necessary.
This includes special handling in the rms and set_sample_width methods of the AudioSegment class.

It changes the codec used for reading from 8-bit wav files to pcm_u8 from pcm_s8
since pcm_u8 is the only valid codec for 8-bit wav files.

pydub/audio_segment.py

@@ -224,6 +224,9 @@ def __init__(self, data=None, *args, **kwargs):
             self.frame_rate = wav_data.sample_rate
             self.frame_width = self.channels * self.sample_width
             self._data = wav_data.raw_data
+            if self.sample_width == 1:
+                # convert from unsigned integers in wav
+                self._data = audioop.bias(self._data, 1, -128)
 
         # Convert 24-bit audio to 32-bit audio.
         # (stdlib audioop and array modules do not support 24-bit data)
@@ -685,7 +688,7 @@ def is_format(f):
             else:
                 bits_per_sample = audio_streams[0]['bits_per_sample']
             if bits_per_sample == 8:
-                acodec = 'pcm_s8'
+                acodec = 'pcm_u8'
             else:
                 acodec = 'pcm_s%dle' % bits_per_sample
 
@@ -804,14 +807,19 @@ def export(self, out_f=None, format='mp3', codec=None, bitrate=None, parameters=
         else:
             data = NamedTemporaryFile(mode="wb", delete=False)
 
+        pcm_for_wav = self._data
+        if self.sample_width == 1:
+            # convert to unsigned integers for wav
+            pcm_for_wav = audioop.bias(self._data, 1, 128)
+
         wave_data = wave.open(data, 'wb')
         wave_data.setnchannels(self.channels)
         wave_data.setsampwidth(self.sample_width)
         wave_data.setframerate(self.frame_rate)
         # For some reason packing the wave header struct with
         # a float in python 2 doesn't throw an exception
         wave_data.setnframes(int(self.frame_count()))
-        wave_data.writeframesraw(self._data)
+        wave_data.writeframesraw(pcm_for_wav)
         wave_data.close()
 
         # for wav files, we're done (wav data is written directly to out_f)
@@ -920,20 +928,12 @@ def set_sample_width(self, sample_width):
         if sample_width == self.sample_width:
             return self
 
-        data = self._data
-
-        if self.sample_width == 1:
-            data = audioop.bias(data, 1, -128)
-
-        if data:
-            data = audioop.lin2lin(data, self.sample_width, sample_width)
-
-        if sample_width == 1:
-            data = audioop.bias(data, 1, 128)
-
         frame_width = self.channels * sample_width
-        return self._spawn(data, overrides={'sample_width': sample_width,
-                                            'frame_width': frame_width})
+
+        return self._spawn(
+            audioop.lin2lin(self._data, self.sample_width, sample_width),
+            overrides={'sample_width': sample_width, 'frame_width': frame_width}
+        )
 
     def set_frame_rate(self, frame_rate):
         if frame_rate == self.frame_rate:
@@ -1009,10 +1009,7 @@ def split_to_mono(self):
 
     @property
     def rms(self):
-        if self.sample_width == 1:
-            return self.set_sample_width(2).rms
-        else:
-            return audioop.rms(self._data, self.sample_width)
+        return audioop.rms(self._data, self.sample_width)
 
     @property
     def dBFS(self):