fix(llm): Fix when use custom embedding providers.

[lx200916]

Description

This pull request updates the embedding request logic in opencontext/llm/llm_client.py to better handle different LLM providers. The main change is switching the order of provider checks and ensuring that multimodal embeddings are used only for the DOUBAO provider.

Embedding provider logic update:

• Changed the provider check to prioritize LLMProvider.DOUBAO and use multimodal_embeddings only for this provider, while other providers (like OPENAI) continue to use the standard embedding API.