Clarify Audio-JEPA training differences from the ICME checkpoint

Add Audio-JEPA results and link the Transformers release

Link pretrained baselines to their Hugging Face model pages

Link comparison-table entries to published Hugging Face models

Recommend BEST-RQ-2.2 and add the project X-ARES comparison table

Release BEST-RQ-2.1-base Transformers encoder from run 67d94qaa

initial commit