feat: add EnhancedNewPolicy that supports enhanced-random-exploration

[Mingle-2012]

• My pull request targets the main branch of Kea.

• I have tested these changes locally on my machine.

This pull request adds a new policy focused on minimizing the count of repeated exploration for random policy.

Changes Include:

1. Add EnhancedNewPolicy with a hyperparameter decay_factor that controls the decay rate of random choices.
2. Set 1s timeout for all LLM-generated events of NewPolicy.
3. Support swipe action for LLM-generated events of NewPolicy.
4. Check the validity of json string obtained from LLM.
5. Useless full-LLM-exploration is also added. This radical feature was tested to be likely to explore more pages within the same event count. (Maybe)