Add `image-text-to-text` task support to `HuggingFaceLocalChatGenerator`

[bilgeyucel]

Is your feature request related to a problem? Please describe.
HuggingFaceLocalChatGenerator component doesn't support image-text-to-text tasks for HF models, which prevents using local LLMs like Qwen/Qwen3.5-4B with Haystack

Describe the solution you'd like
Add the support for image-text-to-text alongside text-generation and text2text-generation tasks

Additional context
Screenshot from Google Colab

[anakin87]

Let's cut a bit of the scope of this issue.

For the moment, we just want that multimodal models (Qwen3.5, Ministral) can generate text instead of rasing errors.

Passing also images is out of scope and tracked in #9672.