Improve docs of model card evaluation data

[osanseviero]

Section https://huggingface.co/docs/hub/model-cards#evaluation-results

[Wauplin]

Related to (internal) slack message (cc @julien-c):

maybe it's worth it improving https://huggingface.co/docs/hub/model-cards#evaluation-results slighlty, which is not currently great?

• the model-index i.E. the only interesting part is missing from the snippet
• the screenshot is ok but the part on the left could be the actual model-index metadata? (here it looks like it's linking from a JSON?)
• maybe mention that the spec is originally by Papers with code?

From @Wauplin:

Also link (and improve) this doc: https://huggingface.co/docs/huggingface_hub/guides/model-cards#include-evaluation-results. huggingface_hub should make it easy to add eval results to an existing model card but it's not well documented.