fix: raise on 401/403 instead of retrying, and fix README's undocumented async-with usage

[Bre77]

Intent

• The README's examples used async with TeslemetryStream(...) as stream:, but the class has never implemented __aenter__/__aexit__ - following the README raises TypeError: ... does not support the asynchronous context manager protocol on the very first line.
  • Fixed the docs rather than adding a context manager: the connection lifecycle here is deliberately listener-driven (AGENTS.md already documents the async_add_listener/auto-close machinery in detail, with dedicated lifecycle tests), and a context manager whose __aenter__ does nothing but return self would just be a second, redundant way to spell the same thing while implying a connect step that isn't actually needed. Rewrote the two affected examples to plain instantiation + async_add_listener, and reworded the Usage section to describe the listener-driven default plus the manual connect()/close()/listen() alternative that was already correctly documented elsewhere in the README.
• aiohttp.ClientResponseError (raised by connect()'s raise_for_status=True) is a subtype of aiohttp.ClientError, so a 401/403 from a bad or revoked access token was falling into __anext__'s generic client-error handler and being retried forever with backoff - the caller saw no events and no error, indistinguishable from a quiet vehicle.
  • __anext__ now special-cases a 401/403 ClientResponseError: it stops the stream and raises a new TeslemetryStreamAuthenticationError (chaining the original error), instead of retrying. Every other ClientError, including other response statuses, keeps the existing backoff-and-reconnect behavior unchanged - verified by tests/test_auth_failure.py, which covers both the 401/403 case and a genuine transient ClientError still retrying and reconnecting.

Ran uv run --with ruff ruff check . and uv run --with mypy mypy teslemetry_stream (both clean save for pre-existing lint findings in the example scripts, unrelated to this change), and every tests/test_*.py script directly.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.