feat: add --agent-mode flag and AUTH0_AGENT_MODE env

[ramya18101]

🔧 Changes

ntroduces Agent Mode — a single switch that optimizes CLI output for AI agents and automation. When enabled, it applies three defaults in PersistentPreRunE, before any subcommand runs: sets --json, --no-input, --no-color`

Decisions:

• Generic scope. Flag, env, and behavior all apply CLI-wide (no per-resource gating).
• Explicit flags always win. Agent mode sets defaults, never overrides — each toggle applies only if that flag wasn't explicitly passed (Cobra flag.Changed). auth0 apps list --csv in agent mode still emits CSV.
• Env beats flag; both beat auto-detect.
• Opt-out is first-class. AUTH0_AGENT_MODE=false (or --agent-mode=false) forces it off even inside a detected agent; surfaced in a stderr notice on activation and in the flag help.
• Notice → stderr, keeping stdout clean for JSON parsing.
• No refactor of the 109 per-command --json registrations. Agent mode rides the shared cli.json field that configureRenderer() already consumes, so output flips without touching per-command flag wiring. Global---json consolidation deferred to a separate ticket.
• --force deliberately excluded — agent mode never auto-confirms destructive operations.

Applies to all commands (generic, CLI-wide).

📚 References

• Builds on the detectAgent() agent-client detection work.

🔬 Testing

Unit (TestResolveAgentMode, table-driven): env-true-over-flag, env-false-over-flag, invalid-env-falls-through-to-flag, flag-true, flag-false.

Manual (live tenant):

AUTH0_AGENT_MODE=true  auth0 apps list          # → JSON
AUTH0_AGENT_MODE=false auth0 apps list          # → table (baseline)
AUTH0_AGENT_MODE=true  auth0 actions list --csv # → CSV (explicit flag wins)
CLAUDECODE=1 AUTH0_AGENT_MODE=false auth0 apps list  # → opt-out honored
CLAUDECODE=1 auth0 apps list                    # → auto-detected on

📝 Checklist

• All new/changed/fixed functionality is covered by tests (or N/A)
• I have added documentation for all new/changed functionality (or N/A)