fix: add codestral-embed model support for OpenAI-compatible provider

[roomote]

Summary

This PR fixes issue #6141 where the codestral-embed model fails with 3072 dimensions when used via the OpenAI-compatible provider.

Problem

When users configure the OpenAI-compatible provider to use the codestral-embed model with 3072 dimensions, the system was unable to find the dimension configuration and would fall back to default behavior, causing a dimension mismatch error in Qdrant.

Solution

Added the codestral-embed model configuration to the openai-compatible provider section in embeddingModels.ts with the correct 3072 dimension support.

Changes

• Added codestral-embed model to the openai-compatible provider with dimension: 3072 and scoreThreshold: 0.4
• Added comprehensive test coverage in embeddingModels.spec.ts to ensure the model configuration works correctly

Testing

• All existing tests pass
• New tests added to verify the model configuration
• Tested with service-factory and config-manager test suites

Fixes #6141

---

Important

Fixes issue #6141 by adding codestral-embed model support with 3072 dimensions to the OpenAI-compatible provider and adds tests for verification.

• Behavior:
  • Fixes issue codestral-embed: Fails with 3072 dimensions via OpenAI-compatible provider #6141 by adding codestral-embed model to openai-compatible provider in embeddingModels.ts with 3072 dimensions.
  • Sets scoreThreshold to 0.4 for codestral-embed.
• Testing:
  • Adds tests in embeddingModels.spec.ts to verify codestral-embed configuration for openai-compatible and mistral providers.
  • Tests getModelDimension() for correct dimension retrieval.
• Misc:
  • Ensures all existing tests pass and adds new tests for model configuration verification.

^{This description was created by for 34e67cf. You can customize this summary. It will automatically update as commits are pushed.}

[daniel-lxs]

Not a fix at all, the issue might not have enough information