Audio encoder tests: Create audio encoders the new way

Specifically, don't expect the ACM to be able to create encoders; we
have to give it an encoder that we make ourselves.

The new way of creating encoders used a 32 kbit/s bitrate
unconditionally for iSAC; I had to change it to 32 kbit/s for 16 kHz
and 56 kbit/s for 32 kHz, which is what the old way of creating
encoders has used since forever.

I also had to change some test expectations on Opus, because the new
way defaults to 32 kbit/s for mono and 64 kbit/s for stereo (which I
believe to be correct), while the old way defaults to 64 kbit/s in
both cases.

Bug: webrtc:8396
Change-Id: I3aab944175a8e27f4c63380e822b27e839bba7f2
Reviewed-on: https://webrtc-review.googlesource.com/94540
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#24375}

modules/audio_coding/acm2/acm_send_test.h

@@ -42,8 +42,9 @@ class AcmSendTestOldApi : public AudioPacketizationCallback,
                      int payload_type,
                      int frame_size_samples);
 
-  // Registers an external send codec. Returns true on success, false otherwise.
-  bool RegisterExternalCodec(AudioEncoder* external_speech_encoder);
+  // Registers an external send codec.
+  void RegisterExternalCodec(
+      std::unique_ptr<AudioEncoder> external_speech_encoder);
 
   // Inherited from PacketSource.
   std::unique_ptr<Packet> NextPacket() override;