Reuse frame decoder for sequential frame retrievel

Allow sequential frame retrieval to reuse the decoder without
re-init and re-seek.

Get rid of IMediaMetadataRetriever api that retrieves multiple
frames at a time, this method could hold up more memory than
what's needed, because the client side has to copy the frames
to a separate set of bitmaps anyways.

This change is needed to support animation Drawbles efficiently
which typically get one frame at a time.

bug: 120414514
test: cts MediaMetadataRetrieverTest, local tests that tests MetadataRetriever thumbnails

Change-Id: I09924408b51bc7491fc5dee121dce9a286c5911f
diff --git a/media/libmediaplayerservice/MetadataRetrieverClient.h b/media/libmediaplayerservice/MetadataRetrieverClient.h
index 272d093..8020441 100644
--- a/media/libmediaplayerservice/MetadataRetrieverClient.h
+++ b/media/libmediaplayerservice/MetadataRetrieverClient.h
@@ -56,9 +56,8 @@
             int index, int colorFormat, bool metaOnly, bool thumbnail);
     virtual sp<IMemory>             getImageRectAtIndex(
             int index, int colorFormat, int left, int top, int right, int bottom);
-    virtual status_t getFrameAtIndex(
-                std::vector<sp<IMemory> > *frames,
-                int frameIndex, int numFrames, int colorFormat, bool metaOnly);
+    virtual sp<IMemory>             getFrameAtIndex(
+            int index, int colorFormat, bool metaOnly);
     virtual sp<IMemory>             extractAlbumArt();
     virtual const char*             extractMetadata(int keyCode);