C-Netz: Calibration/fix processing of tone levels on TX and RX side

FSK and compander levels are adjusted.

Eliminate offsets between subsequent speech chunks. This is done by
high-pass filter. An offset is not passed to the filter.

Do correct audio processing chain:
 time compress -> compressor -> scrambler / pre-emphasis -> TX
 RX -> de-scrambler / de-emphasis -> expander -> time expand

src/cnetz/cnetz.h

@@ -144,8 +144,8 @@ typedef struct cnetz {
 	double			frame_last_phase;	/* master's bit phase of last frame sync */
 
 	/* audio offset removal */
-	double			offset_removal_factor;	/* how much to remove every sample */
-	int16_t			offset_last_sample;	/* last sample of last audio chunk */
+	double			offset_factor;		/* filer alpha of high-pass filter */
+	double			offset_y_last;		/* last stored sample */
 
 	/* measurements */
 	int			measure_speed;		/* measure clock speed */