Add set_timestamp to cma's DecoderBufferBase

chromecast/media/cma/base/decoder_buffer_base.h

Index: chromecast/media/cma/base/decoder_buffer_base.h
diff --git a/chromecast/media/cma/base/decoder_buffer_base.h b/chromecast/media/cma/base/decoder_buffer_base.h
index 98498eb7b534485592093bb408fab9174b46c271..1fca1b27e91276e6313ec79f87eb35883aa7e49d 100644
--- a/chromecast/media/cma/base/decoder_buffer_base.h
+++ b/chromecast/media/cma/base/decoder_buffer_base.h
@@ -33,6 +33,9 @@ class DecoderBufferBase
   // Returns the PTS of the frame.
   virtual base::TimeDelta timestamp() const = 0;
 
+  // Sets the PTS of the frame.
+  virtual void set_timestamp(const base::TimeDelta& timestamp) = 0;

[gunsch, 2015/06/05 20:02:48]

Today, DecoderBufferBase is immutable, which is a pretty nice property. We can
pass every single DecoderBufferBase around as const.

What about making something like a builder that makes a copy of a
DecoderBufferBase with a new timestamp, or perhaps a subclass
(TimeShiftedDecoderBufferBase) that takes a DecoderBufferBase and a new
timestamp in its constructor?

[halliwell, 2015/06/05 20:16:47]

Hmmm ... I could definitely do either of those if we want these to be immutable,
although:
* Although it's all 'const', it's not exactly immutable (see writable_data!)
* I don't believe we pass them around as const.  They seem to be passed as const
scoped_refptr<_>& a lot, which is fine to call non-const members on the enclosed
pointer.

Let me know what you think :)

+
   // Gets the frame data.
   virtual const uint8* data() const = 0;
   virtual uint8* writable_data() const = 0;