Redesign chat creation modal

[pmeier]

When starting a new chat in the UI the controls in the modal can be grouped into three categories:

• 🟦 Generic parameters, e.g. chat name, or controls, e.g. starting the chat or cancelling the creation.
• 🟩 Parameters related to the preparation stage of RAG, e.g. selecting source storage and parametrizing the chunking. In the light of "managed" Ragna #256, uploading documents also falls into this category, since uploading documents will only be our default use case
• 🟥 Parameters related to the interrogation stage of RAG, e.g. selecting an assistant and parametrizing the source retrieval.

This is how it currently looks like

All categories are mixed together. To make progress on #256, we need to cleanly separate the three in order for 🟩 and 🟥 to become pluggable components.

While creating a design for this, we should also keep #191 and #217 in mind. TL;DR the amount of controls we currently have is going to grow. For 🟩 we are likely going to get at least one more drop down menu for the embedding model to use. And for all components, i.e. embedding model, source storage, assistant, we are going to have an unknown amount of controls that depend on what the user defines in the configuration.

[smeragoel]

Based on discussion with @pmeier, here are the new chat creation screen designs:

Figma Links:

1. Screens
2. Interactive Prototype

PNGs: