Use PIL Image internally for the Multimodal Agent

[BeibinLi]

Why are these changes needed?

As many people have observed, saving a base64 string in _oai_messages would make debugging difficult because the string is so long that it would spam the terminal or output log (e.g., #1087). Therefore, we save a 'PIL image' inside _oai_messages and convert it to base64 before calling the OpenAI client.

Related issue number

Checks

• I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://microsoft.github.io/autogen/docs/Contribute#documentation to build and test documentation locally.
• I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.

[codecov-commenter]

Codecov Report

Attention: 13 lines in your changes are missing coverage. Please review.

Comparison is base (9708058) 39.46% compared to head (d0e49b9) 50.99%.

Files	Patch %	Lines
.../agentchat/contrib/multimodal_conversable_agent.py	59.09%	8 Missing and 1 partial ⚠️
autogen/agentchat/contrib/img_utils.py	93.18%	1 Missing and 2 partials ⚠️
autogen/agentchat/contrib/llava_agent.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #1124       +/-   ##
===========================================
+ Coverage   39.46%   50.99%   +11.52%     
===========================================
  Files          57       57               
  Lines        6020     6073       +53     
  Branches     1346     1478      +132     
===========================================
+ Hits         2376     3097      +721     
+ Misses       3449     2726      -723     
- Partials      195      250       +55     

Flag	Coverage Δ
unittests	50.91% <80.59%> (+11.44%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.