Add support for @model for LLM models

[joulei]

Context

For the agent integration with MoFa, we've added a @name selector capability within the chat input prompt that allows the user to select different agents. See https://github.com/moxin-org/moly/pull/279.

We now want to include this behaviour for LLM models within a chat.

Requirements

• Writing @ should bring up a list of models to select from (it should match the items and status of the model selector at the top of the chat panel.
• The user should be able to navigate the list via keyboard and select a model.
• Once the model is selected the UI should show the selection above the message input (see screenshot above).
• Sending a message with the newly selected model will load the model first and then send the message. Once the message is sent the text input should be reset to its initial state.
• The user should not be able to select the already current model, that item in the list should be shown as already selected.

[joulei]

We should actually wait to work on this once https://github.com/moxin-org/moly/pull/279 is merged. I just realized the feature is in the same input prompt, let's avoid more conflicts and add @model support to the already existing feature in that branch.