Avoiding Cloud Lock-in

@jordanstephens and I were talking and I kinda feel like we should aim to write the docs in such a way that we guide the users to avoid vendor lock-in.

Specifically I was thinking about this page:
https://docs.defang.io/docs/concepts/managed-llms/managed-language-models

Technically, this example is showing people how to set themselves up for vendor lock-in: we're showing them how to build an app that will enable the necessary features in their cloud so that they can use the cloud-specific APIs.

Personally, I think we should flip that page to be more like this page:
https://docs.defang.io/docs/concepts/managed-llms/openai-access-gateway

i.e. by default we should encourage people to use the option that is more portable.

I think there may be a lot of value in the APIs opened up by Bedrock/Vertex without being forced through the OpenAI API, so I think it's fair to keep that on the page, but perhaps at the bottom with a prominent message/caveat saying using this feature (x-defang-llm) without that gateway will lock you into your current cloud provider.

Start with the Managed Language Models page and compare it with the OpenAI Access Gateway page, especially their examples. Update the guidance so the portable gateway option is the default, while retaining the Bedrock/Vertex-specific x-defang-llm option with a prominent cloud lock-in caveat; done means the page clearly presents both paths and their portability trade-offs.