LibVideo/VP9: Retain adjacent block contexts storage between frames

Re-allocating the storage is unnecessary, since the size will rarely
change during playback.

Userland/Libraries/LibVideo/VP9/Context.h

@@ -253,6 +253,11 @@ struct ColorConfig {
 
 struct FrameContext {
 public:
+    FrameContext(Vector2D<FrameBlockContext>& contexts)
+        : m_block_contexts(contexts)
+    {
+    }
+
     u8 profile { 0 };
 
     FrameType type { FrameType::KeyFrame };
@@ -338,7 +343,7 @@ private:
     //        arrays instead.
     //        I think should also apply to other fields that are only accessed relative to the current block. Worth looking
     //        into how much of this context needs to be stored for the whole frame vs a row or column from the current tile.
-    Vector2D<FrameBlockContext> m_block_contexts;
+    Vector2D<FrameBlockContext>& m_block_contexts;
 };
 
 struct TileContext {

Userland/Libraries/LibVideo/VP9/Parser.cpp

@@ -147,7 +147,10 @@ DecoderErrorOr<ColorRange> Parser::read_color_range()
 /* (6.2) */
 DecoderErrorOr<FrameContext> Parser::uncompressed_header()
 {
-    FrameContext frame_context;
+    // NOTE: m_reusable_frame_block_contexts does not need to retain any data between frame decodes.
+    //       This is only stored so that we don't need to allocate a frame's block contexts on each
+    //       call to this function, since it will rarely change sizes.
+    FrameContext frame_context { m_reusable_frame_block_contexts };
     frame_context.color_config = m_previous_color_config;
 
     auto frame_marker = TRY_READ(m_bit_stream->read_bits(2));

Userland/Libraries/LibVideo/VP9/Parser.h

@@ -175,6 +175,7 @@ private:
     ReferenceFramePair m_comp_var_ref;
 
     bool m_use_prev_frame_mvs;
+    Vector2D<FrameBlockContext> m_reusable_frame_block_contexts;
     Vector2D<PersistentBlockContext> m_previous_block_contexts;
     // Indexed by ReferenceFrame enum.
     u8 m_mode_context[4] { INVALID_CASE };