Add CLAUDE.md with development setup instructions

[max-sixty]

Summary

• Add CLAUDE.md file with essential development commands for xarray
• Includes setup with uv sync, test commands, and linting instructions

Test plan

• uv sync runs successfully
• uv run pytest xarray -n auto executes tests
• pre-commit run --all-files runs linting checks
• uv run dmypy run performs type checking

🤖 Generated with Claude Code

[max-sixty]

are folks OK with this?

two decisions:

• is it OK to have Claude Code specific files with instructions?
• is it OK to suggest to the agents that it use uv

I'd like to merge this...
...but reasonable options from here:

• have other files like AGENTS.md which are symlinked
• additionally add miniconda instructions

[TomNicholas]

I think these text files represent low risk and low commitment to us, so you (and I know @ianhi is interested) should just iterate how you like.

Running AI agents is always at your own risk, I don't think we could be reasonably blamed for writing something sensible but opinionated in a text file in our repo.

I would pause only before changes to actual user-facing docs - those should be human-prioritizing.

[max-sixty]

yes, tbc, folks are definitely responsible for anything that's merged from them, whether they typed the words or an AI agent did!

I also hold a norm that we should all be transparent when an AI agent does work — both in code or comments — we shouldn't have AIs impersonating us. it's fine to say "it's the AI's words. I stand behind them and take responsibility for them but an AI wrote them"

[TomNicholas]

Agree - might want to add something like that to contributor's guidelines and PR template.