Refactor model-specific configs and move data curation scripts

[SumanthRH]

What does this PR do?

Another refactor PR on top of #23 now focused on model-specific configurations and data generation.

• Model-specific system prompts, user templates etc are best left to be in the a YAML file.
• TaskHandler should be model agnostic, since we want to have a consistent evaluation logic for all tasks
• Data curation scripts for different Sky-T1 models should live outside the skythought_evals package. These are mostly scripts focused on a particular data curation task like filtering, rewriting etc. My proposal is to place common scripts in scripts/ . A guide for obtaining the final training data + training commands for different Sky-T1 models should be placed in recipes/ .

scripts can be organized better, but for now I've placed all the data curation scripts in this folder.

TODO:

• Move Sky-T1 specific instructions to recipes/
• Improve READMEs -skythought_evals/README.md and README.md
• Verify with E2E tests for Sky-T1

[SumanthRH]

This PR is now ready for review.

Since i've moved the commands for data curation, it would be good for a dummy test run on small data for these. I don't have the intermediate data for some of the commands (like convert_data.py etc) so it would be good to test this out @caoshiyi @lynnliu030 @tyler-griggs

[SumanthRH]

does this look good? @tyler-griggs

[tyler-griggs]

This is fantastic. Thanks!

[lynnliu030]

Thanks for the work!