Pass system message as system_instruction to Gemini #1011.

[liminma]

@rjambrecic

Why are these changes needed?

Gemini supports system instructions. Sending the system_message as a system_instruction could potentially be more effective.

Related issue number

Fixes #1011

Checks

• I've included any doc changes needed for https://docs.ag2.ai/. See https://docs.ag2.ai/docs/contributor-guide/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.

[CLAassistant]

All committers have signed the CLA.

[sonichi]

Thanks for the contribution. Thanks to this PR, I found some issues with the current documentation in https://docs.ag2.ai/docs/user-guide/models/google-gemini.

[marklysze]

Thanks @liminma!

[codecov]

Codecov Report

Attention: Patch coverage is 5.26316% with 18 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
autogen/oai/gemini.py	5.26%	18 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (0790aae) and HEAD (154b59b). Click for more details.

HEAD has 1273 uploads less than BASE

Flag	BASE (0790aae)	HEAD (154b59b)
core-without-llm	15	1
3.9	78	0
ubuntu-latest	147	1
3.13	88	0
macos-latest	108	0
3.11	67	1
3.10	97	0
commsagent-discord	9	0
optional-deps	139	0
3.12	39	0
browser-use	7	0
windows-latest	114	0
commsagent-slack	9	0
commsagent-telegram	9	0
crawl4ai	13	0
docs	6	0
graph-rag-falkor-db	6	0
interop	13	0
interop-crewai	9	0
interop-langchain	9	0
interop-pydantic-ai	9	0
jupyter-executor	9	0
twilio	9	0
retrievechat	10	0
retrievechat-pgvector	10	0
websockets	9	0
retrievechat-mongodb	10	0
retrievechat-qdrant	11	0
agent-eval	1	0
gpt-assistant-agent	3	0
lmm	4	0
teachable	4	0
gemini	15	0
websurfer	15	0
anthropic	16	0
bedrock	14	0
cerebras	14	0
llama-index-agent	3	0
retrievechat-couchbase	3	0
cohere	15	0
groq	14	0
ollama	14	0
mistral	14	0
together	14	0
swarm	14	0
reasoning	14	0
long-context	3	0
integration	12	0
falkordb	2	0
autobuild	1	0
core-llm	4	0
neo4j	2	0
deepseek	1	0
openai	1	0

Files with missing lines	Coverage Δ
autogen/oai/gemini.py	13.01% <5.26%> (-46.94%)	⬇️

... and 62 files with indirect coverage changes

[liminma]

Thanks @liminma!

Glad to have contributed to the project! Thanks for merging.