feat(litellm): align embedding instrumentation with pending spec

[codefromthecrypt]

This PR aligns litellm with the specification changes around embeddings in #2162

Spec changes from 2162

• Consistent span name: "CreateEmbeddings"
• Standardized attribute structure: embedding.embeddings.N.embedding.{text|vector}
• Unified invocation parameter tracking: embedding.invocation_parameters
• Proper llm.system attribute for provider identification

Code improvements:

• Full batch embedding support with indexed attributes
• Separated invocation parameters from input data
• Improved handling of token IDs vs text inputs
• Vectors stored as tuples instead of JSON strings

This is the same as #2210, except litellm.

---

Note

Aligns LiteLLM embedding spans to the new spec with "CreateEmbeddings" name, structured per-embedding text/vector, invocation parameter capture, provider tagging, and full batch handling.

• Embedding Instrumentation (LiteLLM):
  • Rename span to CreateEmbeddings; set llm.system to "litellm".
  • Capture invocation params under embedding.invocation_parameters (exclude input).
  • Record inputs as embedding.embeddings.N.embedding.text for strings/lists of strings; skip for token IDs.
  • Emit vectors as tuples per item at embedding.embeddings.N.embedding.vector.
• Result Handling:
  • Iterate EmbeddingResponse.data to set per-index vectors instead of a single JSON string.
• Tests:
  • Add batch and edge-case embedding tests (tests/test_batch_embedding.py).
  • Update existing tests to expect new span name, attributes, vectors-as-tuples, and invocation parameters.
  • Add setup_litellm_instrumentation fixture.

^{Written by Cursor Bugbot for commit c0c6750. This will update automatically on new commits. Configure here.}