Fix incorrectly parsing event fieldnames in yielding responses

[idantene]

• I understand that this repository is auto-generated and my pull request may not be merged

Changes being requested

Remove requirement for event names to end with a dot if they're "response" or "transcript", and extend the notion of a "valid" event name to be either missing, response, transcript, or an extended of either (response. and transcript.), with separate logic to reduce code duplicity.

Additional context & links

When parsing the SSE in a streaming request, the captured event (if any) is parsed to ensure it is valid/matches the expected event.
It seems there are three options for a "valid event" - a missing one, and either a response or transcript one. In the current implementation, the latter are forced to have a trailing dot (supposedly to offer more granular information on the event type, such as response.foo), but this is out of scope for SSEs.
In fact, a server yielding event:response gets the wrong path and ends up with a response chunk of process_data(data={"data": data, "event": sse.event}, cast_to=cast_to, response=response), which is then incorrectly parsed on the client side.