Fix Image.frombytes() mode param usage

[akhilam512]

currently we're passing frame.format which typically contains file format strings like 'PNG' or 'JPEG' as the mode parameter to PIL.Image.frombytes()

This function actually expects image modes like 'RGB', 'RGBA', etc

I standardized all Image.frombytes() calls to use 'RGB' mode instead, fixed this in six files:

• src/pipecat/transports/base_output.py
• src/pipecat/services/gemini_multimodal_live/events.py
• src/pipecat/services/moondream.py
• src/pipecat/services/google/google.py
• src/pipecat/services/anthropic.py
• src/pipecat/processors/aggregators/openai_llm_context.py

For most of the places where it's used, 'RGB' as the image mode works perfectly since it is immediately convert to JPEG anyway. However in base_output.py, I went with 'RGB' for consistency, though I wonder if preserving other modes is useful? Not sure if a png with an alpha channel can be passed here for example, and we might prefer RGBA in that case (to account for that, we might need to add a dedicated class member for storing image mode in ImageRawFrame perhaps)

[akhilam512]

Hmm sorry this isn't a complete solution still, I was a bit too eager to send the PR. Testing this now I get:

unexpected exception: not enough image data

I think this is because as its name suggests frombytes() expects raw bytes but we're still sending a compressed bytes via a png/jpg file. So frombytes() is probably not the right usage here at all, instead using something like Image.open(io.BytesIO(image)) and then saving the resultant image seems to work

[markbackman]

@akhilam512 mind if I turn this into a draft until it's ready for review?

[markbackman]

@akhilam512 is this something you're still working on or can I close it?