Add unified client [DRAFT]

[habara-k]

Motivation

The current v3 SDK is auto-generated from the LINE OpenAPI specs, where each YAML file produces a separate client class (e.g., MessagingApi, ManageAudience, Insight, etc.). This means users must instantiate and manage multiple clients individually, leading to verbose boilerplate:

from linebot.v3.messaging import Configuration, ApiClient, MessagingApi
from linebot.v3.insight import Configuration as InsightConfiguration, ApiClient as InsightApiClient, Insight

messaging_config = Configuration(access_token='TOKEN')
insight_config = InsightConfiguration(access_token='TOKEN')

with ApiClient(messaging_config) as api_client, InsightApiClient(insight_config) as insight_client:
    messaging_api = MessagingApi(api_client)
    insight_api = Insight(insight_client)
    ...

This PR introduces a unified client that wraps all per-domain clients behind a single entry point:

from linebot.v3 import LineBotClient

with LineBotClient(channel_access_token='TOKEN') as client:
    client.push_message(...)
    client.get_number_of_followers(...)

What's changed

• Add tools/generate_unified_client.py — a script that introspects the per-domain API classes and generates LineBotClient / AsyncLineBotClient wrapper classes. This script is hooked into generate-code.py so it runs automatically after the OpenAPI generator completes.
• Add LineBotClient and AsyncLineBotClient (auto-generated) under linebot.v3
• Update all examples (flask-echo, flask-kitchensink, aiohttp-echo, fastapi-echo, rich-menu, simple-server-echo) and README.rst to use the unified client
• Add MIGRATION.md — comprehensive migration guide from legacy linebot (v2) to linebot.v3, covering client construction, imports, method mapping, error handling, and a full before/after echo-bot example
• Deprecate all legacy linebot.models and linebot.http_client classes with @deprecated decorator pointing users to linebot.v3 equivalents
• Add Sphinx doc generation for v3 modules
• Add tests for both sync and async unified clients

Design decisions

async_req parameter is intentionally excluded

The OpenAPI-generated per-domain clients (e.g., AsyncMessagingApi) support an async_req parameter on every method. When async_req=True, the request runs in a thread pool and returns an AsyncResult:

thread = messaging_api.push_message(req, async_req=True)
result = thread.get()

This is a thread-based concurrency escape hatch for users of the sync client who want non-blocking behavior without asyncio. The unified clients intentionally drop this parameter because neither use case needs it:

• LineBotClient — designed for straightforward synchronous usage. Users who need thread-based concurrency can access the underlying per-domain clients directly (e.g., client._messaging_api).
• AsyncLineBotClient — already fully async via asyncio/aiohttp. The thread pool path would be counterproductive.

How to review

• Doc: Build and preview locally:

  cd docs && make html && open build/html/index.html