[SPARK-56449][INFRA] Skip pre-flight checks for code questions in AGENTS.md

[szehon-ho]

What changes were proposed in this pull request?

Updated AGENTS.md to skip the pre-flight checks when the user is only asking questions about the code or its behavior.

Why are the changes needed?

For every new AI agent session (cursor, claude code), a new questions like "what does function X do?" previously triggered git remote -v, a staleness check, a potential git fetch, git status, and a branch switch.

It's somewhat of a regression from the previous, more lightweight experience before AGENTS.md. It happens whenever open a tabs to start a different question session about the code.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

Manual verification — this is a documentation-only change to AI agent instructions.

Was this patch authored or co-authored using generative AI tooling?

Generated-by: Cursor (Claude Opus 4)

[szehon-ho]

Hi @cloud-fan , i noticed this is a bit of a regression for open a new tab and asking another question about the code (it does git fetch every time and potentially rebase), which is one of my main use case (though maybe its not the main use case for you or other experienced Spark devs :)) . wdyt?