Support for Multiple Embedding Model Providers (e.g., Gemini)

### Required Pre-requisites

- [x] I have read the [Documentation](https://www.aci.dev/docs)
- [x] I have searched the [Issue Tracker](https://github.com/aipotheosis-labs/aci/issues) and [Discussions](https://github.com/aipotheosis-labs/aci/discussions) that this hasn't been reported yet.
- [ ] Consider asking in [Discussions](https://github.com/aipotheosis-labs/aci/discussions) first

### Motivation

Currently, the ACI.dev platform MCP search function appears to rely exclusively on OpenAI's embedding models (specifically text-embedding-3-small) for function search and vector similarity matching. This dependency might require users to provide an OpenAI API key to use the search functionality.

To potentially increase system flexibility and reduce dependency on a single vendor, it might be beneficial to add support for alternative embedding model providers, such as Google's embedding models.

### Proposed Solution

It might be possible to modify the system to support multiple embedding model providers:

Add new environment variables like SERVER_EMBEDDING_PROVIDER
Add provider-specific configuration options
Create an abstraction layer for embedding services that could support different implementations

Modify the embedding generation logic to use the appropriate service based on configuration

Ensure vector dimensions and similarity calculation methods remain compatible with existing database embeddings