Add support for remote embeddings providers, additional fixes

[hgueorguiev]

• Add support to configure if embeddings are local or use remote service to gen.
• Add support for HuggingFace inference endpoints for embedding generation
• Fix missing torch lib wheel that poetry fails to install on linux, this allows us to build a working docker container image again
• Fix missing weaviate dependency by adding to project requirements
• File preview route to be more forgiving of files it fails to generate a preview for load whatever files it is able to support

[vercel]

Someone is attempting to deploy a commit to the Sidekick-ai Team on Vercel.

A member of the Team first needs to authorize it.

[hgueorguiev]

@Ayan-Bandyopadhyay Let me know how you feel about these changes, mainly abstracting embeddings providers so that they can be processed remotely on gpu hardware, first use case HuggingFace inference points. But also some fixes for issues I was running into.