feat: per-chat model configuration override

[cezarlugullo]

Summary

This PR adds the ability for users to set a custom model for individual chats without changing the global settings.

Problem

Currently, changing the chat model affects all conversations. Users who want to test different models or use a specific model for a particular task must change the global settings back and forth.

Solution

A per-chat model override that lets users toggle a custom model configuration on any chat:

• Toggle on: the chat uses its own provider, model name, and parameters
• Toggle off: the chat reverts to the global settings
• The override is persisted and automatically re-applied when the chat is reloaded from disk

How it works

Backend

• helpers/chat_model_override.py — core logic: get_override(), set_override(), apply_override()
• api/chat_model_get.py — GET endpoint returning current model config (global or custom)
• api/chat_model_set.py — SET endpoint to enable/disable/update the override
• helpers/persist_chat.py — 4 lines added to re-apply override on chat deserialization

Frontend

• A tune button (🎵) next to each chat in the sidebar
• Opens a modal where users can:
  • Toggle custom model on/off
  • Select provider and model name
  • Configure API base, context length, vision support, rate limits, and extra kwargs
• Built with Alpine.js following the existing modal pattern (x-data store, dynamic import)

Tests

• tests/test_chat_model_per_chat_override.py — unit tests covering:
  • Setting/removing override data
  • Type normalization (string numbers, booleans, kwargs parsing)
  • Fallback behavior when fields are empty or invalid

Screenshots

The modal follows the existing UI style. A small tune icon appears next to each chat name in the sidebar.

Checklist

• Feature branch based on development
• Follows existing code patterns (ApiHandler, Alpine.js stores, modal structure)
• Includes unit tests
• No unrelated changes