Update docs with benchmark tutorial

[brunosan]

After #117 has closed, I think we should update the documentation in two places, which I think are both small changes:

• Add a new model benchmark section, so in all future versions we compared how the scores are improving.
• Add the code in the finetune section to help users apply finetuned for a random places with random labeled data, using the floods as the example.

[lauracchen]

Re-upping this @yellowcap - can we incorporate this in the next sprint?

[yellowcap]

Potentially, we did work on this back in December and early January, but without having a finalized version. In January we de-prioritized the systematic benchmarking in favor of creating more hands-on examples.

@lillythomas what would it take to document how to do fine tuning?

Refs #87 #117