Add AGENTS.md for AI coding agents

[riverma]

Adds an AGENTS.md at the repo root as the agent-facing companion to CONTRIBUTING.md and NOMENCLATURE.md. It summarizes the nine CI lint gates from CL.yml with the exact commands and the most common pitfalls, ships a copy-pasteable Python file skeleton matching the house style in _rapid2.py, and condenses the naming grammar to a scannable quick reference pointing back to NOMENCLATURE.md for the full tables. Goal is that a coding assistant with this file in context produces a PR that passes lint on the first try.

Also adds CLAUDE.md as a symlink to AGENTS.md so Claude Code (which reads CLAUDE.md) and other tools (which read AGENTS.md) share one source of truth.

Closes #40.

[riverma]

Quick note on how this was verified locally before opening the PR.

Spawned a fresh Sonnet coding agent in an isolated worktree with the new AGENTS.md as its only project-side guidance, and gave it a small task: "add a -vrb / --verbose flag to _rapid2.py following project conventions." The agent picked the right short-flag convention and correctly named the parsed value BS_vrb (boolean + scalar, derived from the nomenclature section). Of the nine CI lint commands run on the agent's diff, seven passed cleanly. The other two non-passes don't reflect on AGENTS.md or the agent's work: shellcheck wasn't installed in my local environment (and the agent didn't touch any shell scripts, so it wouldn't have fired anyway), and mypy errored on a third-party earthaccess import that uses Python 3.12 syntax incompatible with the project's python_version = "3.11" setting — the same failure happens on a clean checkout of main.

The agent flagged one gap: no explicit worked example for naming boolean CLI flags. Added the BS_vrb / BS_<short-flag-name> example to Section E to close that.

[c-h-david]

Following up on our conversation yesterday... Would you consider minimizing duplication with other files and instead pointing to them such that maintenance is eased? I understand this might have an impact on "tokens" so it's worth acknowledging here. Notably, the file TESTING.md has much of the description of linting, formatting and testing. Also pyproj.toml includes important linting and formatting options, so do .pymarkdown.yml and .yamllint.yml. I do really like the CLI formatting standard and I wonder if that could justify a separate file of its own pointed to here. I also like the part on registration of CLIs into pyproj.toml. Actually I feel like the pre-PR checklist, the authoritative sources, and the formatting guidance for program structure are the most important new addition and the rest could be perceived as redundant/duplicate/more-maintenance. Thoughts?