Add human feedback

Use human feedback for demonstration data, currently there is only one strategy: simple which asks for interaction until the maximum is reached.

There are some new and changed CLI parameters:

• --evolution-demo-source instead of --use-evolution-demo takes 1+ arguments (i.e., both can be used): static (using predefined evolution demos) and interactive (using human feedback)
• --annotation-strategy (currently only simple (always use human feedback until maximum, and use all annotations for demonstration) exists)
• --max-annotations: the maximum number of human interactions
• --max-annotations-per-evolution: the idea is to set a limit for interactions per evolution , but this is not used yet

requested review from @griesshaber

added 1 commit

• 0e697c53 - Use default value for annotation strategy CLI parameter

Compare with previous version

changed the description

added 1 commit

• c203cb79 - Consider limit for annotations per evolution

Compare with previous version

added 1 commit

• 1205a889 - Update default value for CLI parameter `--max-annotations-per-evolution`

Compare with previous version

changed the description

added 2 commits

• fe66ab5a - Log user interactions
• 91e60332 - Update response editor to better show response

Compare with previous version

added 1 commit

• 80057a95 - Refactor evolution and add GA-CoT

Compare with previous version

added 1 commit

• 4293bbea - Add script for downloading wandb experiments

Compare with previous version

mentioned in commit 02102671

merged