Updates to the ChatGenerator protocol

Two ideas for updating the ChatGenerator protocol:

1. Update the return type of the run method to at least require replies: list[ChatMessage] since we rely on that key being present when we use the protocol in the codebase. This should be possible through using TypedDict, total=False and Required (still needs to be tested). So something like

from typing import Protocol
from typing_extensions import Required, TypedDict
from haystack.dataclasses import ChatMessage


class ChatGeneratorOutput(TypedDict, total=False):
    """
    Return type for Chat Generators.

    ``replies`` is always required. Implementations may include additional keys.
    """
    replies: Required[list[ChatMessage]]


class ChatGenerator(Protocol):
    def run(self, messages: list[ChatMessage]) -> ChatGeneratorOutput:

2. Consider adding a run_async to the protocol since we have finished adding run_async to all of our chat generators in core and in core-integrations https://github.com/deepset-ai/haystack/issues/8858 and https://github.com/deepset-ai/haystack-core-integrations/issues/1379. Only a consideration since it's not guaranteed that community made components follow the protocol, but we are starting to add run_async methods to components that use the ChatGenerator protocol so I wonder if creating a second protocol (e.g. ChatGeneratorWithAsync) would be an alternative.

👋 Hello there! This issue will be handled internally and isn't open for external contributions. If you'd like to contribute, please take a look at issues labeled contributions welcome or good first issue. We'd really appreciate it!

Start by locating the ChatGenerator protocol and its implementations, then review how run return values and run_async methods are used across the codebase. Check issues #8858 and #1379 for the existing async work. Done would mean an agreed protocol design with its typing and async implications addressed, though the issue states that this work is handled internally.