[Feature Request]: change `JinaEmbeddingFunction` to support jina-embeddings-v3

[bwanglzu]

Describe the problem

Hi, i was randomly browsing Chroma documentation, realised that the default embedding funfction (API integration) for JinaAI is still jina-embeddings-v2, can we update the default model to support jina-embeddings-v3? The advantage:

1. better performance
2. multilingual (89 languages)
3. MRL embedding

This might need to introduce new parameters:

1. task: choose lora adapter for specific downstream task for optimal performance.
2. dimensions: cut the embedding to maintain good performance while reduce storage
3. late_chunking: get token embeddings, chunk token embeddings then apply pooling for better context-aware.

note: i work for Jina AI.

Additional reading:

1. release blog post: https://jina.ai/news/jina-embeddings-v3-a-frontier-multilingual-embedding-model/
2. paper: https://arxiv.org/abs/2409.10173
3. late chunking: https://jina.ai/news/late-chunking-in-long-context-embedding-models/
4. late chunking paper: https://arxiv.org/abs/2409.04701

Describe the proposed solution

1. replace jina-embeddings-v2 with jina-embeddings-v3.
2. introduce additional parameters
3. update documentation
4. integration test

Alternatives considered

No response

Importance

nice to have

Additional Information

No response