[Performance] Slow loading of MM datas

In `_load_samples_to_ram` and calls `formulate_loaded_sample`. <img width="938" height="172" alt="Image" src="https://github.com/user-attachments/assets/cfe74ceb-bca7-4d54-8f21-ac39e93df4b8" /> Parallelize with `ThreadPoolExecutor`. For the Shopify benchmark, it can take up to 240s, which is unnecessary. Also, it should use pybase64 for encode and decode