Add Anthropic Models to Cache Prompt

[ColeMurray]

Short description of the problem this fixes or functionality that this introduces. This may be used for the CHANGELOG
Adds support for Anthropic models with 'anthropic/' prefix in cache prompting.

---

Give a summary of what the PR does, explaining any non-trivial design decisions
This pull request addresses an issue where Anthropic models with the 'anthropic/' prefix were not being correctly identified for cache prompting support. The changes include:

1. Added two new entries to the cache_prompting_supported_models list:

   • 'anthropic/claude-3-5-sonnet-20240620'
   • 'anthropic/claude-3-haiku-20240307'

2. Retained the existing entries without the 'anthropic/' prefix to maintain backwards compatibility:

   • 'claude-3-5-sonnet-20240620'
   • 'claude-3-haiku-20240307'

This change ensures that the cache prompting feature correctly supports Anthropic models regardless of whether they are referenced with or without the 'anthropic/' prefix. This dual-entry approach allows for flexibility in model naming conventions while maintaining support for existing configurations.

The decision to keep both prefixed and non-prefixed versions was made to:

1. Support users who may be using the models with the 'anthropic/' prefix in their configurations.
2. Maintain compatibility for users who are already using the non-prefixed versions.

---

Link of any specific issues this addresses
Fixes #3774: Incorrect identification of Anthropic models in prompt cache supported models

[enyst]

Thanks for the fix!

This list we have will be a pain to keep updated... 😅 I hope liteLLM will support this feature in a more generic way in the future. For now it's good to have it working.

[tobitege]

This shouldn't be necessary, i.e. the model name should be compared to be in the array entries' name (not equality), so we don't need to add prefixed versions.
Please apply a change to is_caching_prompt_active in llm.py, e.g.

    return (
        self.config.caching_prompt is True
        and any(model in self.config.model for model in cache_prompting_supported_models)
    )

[tobitege]

Please see my comment above.

[tobitege]

LGTM
Thank you for your contribution! 👍