Add support for docker-sandboxed agentic development with Gemini

[jbcoe]

Add agentic development sandbox and consolidate documentation

[philipcraig]

I'm happy to approve as is. For potential future work, I often comment at work that everything related to how to build and test should be in Contributing.md for humans already, and the agent file should just link to that. Also, if Gemini would read an agents.md file, then that would be more general and open to others with different agents than a Gemini specific file

[jbcoe]

@philipcraig I've re-jigged docs to follow your guidance.

[philipcraig]

Lgtm. Are you sure Gemini would not read Agents.md if it were present?

[jbcoe]

@philipcraig Gemini says:

"I don't automatically read AGENTS.md when I start up (my system instructions primarily look for a GEMINI.md file for foundational mandates). However, I noticed it in your directory structure and just took the initiative to read it!

I see that it contains specific instructions for AI agents operating in this workspace, including the mandate to read README.md and CONTRIBUTING.md before executing any tasks, as well as rules for development conventions, git usage, and validation (ensuring CMake and Bazel tests pass).

Would you like me to go ahead and read README.md and CONTRIBUTING.md now to get fully up to speed?"

I'll stick with GEMINI.md until the "standard" becomes so.

[philipcraig]

Lgtm

[nbx8]

All good - thanks!