PMM-7 - Add AI agents development guide

[ondrejpatocka]

Summary

• Add AGENTS.md — a comprehensive development guide for AI coding agents, documenting the Percona/PMM delta over upstream Grafana (architecture, directory structure, state management, API layer, conventions, configuration, dev workflow, and testing).
• Add thin entry-point shims so each major AI assistant auto-discovers the guide through its native convention:
  • CLAUDE.md — Claude Code auto-discovery
  • .cursorrules — Cursor auto-discovery
  • .github/copilot-instructions.md — GitHub Copilot auto-discovery
• Include a self-maintenance policy ("Keeping This Document Current") that instructs agents to update AGENTS.md whenever they add or change PMM features, routes, plugins, Redux slices, API services, CI configs, or environment variables — preventing the guide from going stale over time.

Motivation

AI agents (Claude Code, Cursor, GitHub Copilot, etc.) lack awareness of the Percona fork's conventions, integration points, and // @PERCONA tagging discipline. Without explicit guidance they tend to scatter PMM code across upstream Grafana directories, miss existing API patterns, or break rebase-friendly conventions. A single authoritative document — automatically loaded by each tool — gives every agent session the right context from the start.

What's included

File	Purpose
AGENTS.md	Full guide: PMM context, fork rules, architecture tables (frontend/backend/CI), directory tree, state management, API layer, do/don't conventions, configuration, dev workflow, testing, key file references
CLAUDE.md	Shim — redirects Claude Code to AGENTS.md, repeats PMM context
.cursorrules	Shim — redirects Cursor to AGENTS.md, repeats PMM context
.github/copilot-instructions.md	Shim — redirects GitHub Copilot to AGENTS.md, repeats PMM context

[matejkubinec]

@ondrejpatocka since grafana has their own agents.md, I think it would be good to have a AGENTS.PERCONA.MD (or similar approach) to not to cause issues during pulls.