Andrey Kleshchev authored May 26, 2022 and Rye Mutt committed May 28, 2022

Image thread doesn't need mBuffer and buffer isn't thread safe so no point allocating it in an image thread.

Replace std::map with hashmaps for mesh_load_map and skin_load_map in LLMeshRepo to reduce find overhead

Use new non-allocating unzip and unpack functions to reduce mesh fetch memory and allocation overhead