refactor: improve z.enum support in CLI framework

[hsablonniere]

Context

Enum values defined in z.enum() schemas were manually duplicated in descriptions across 7 files, creating maintenance burden and fragility. When adding a new enum value, every copy had to be updated or things would silently diverge.

Additionally, explicit autocomplete arrays and manual validation were duplicated for each enum, preventing a unified approach to displaying and validating enum choices.

Proposal

Add framework-level introspection for z.enum() schemas:

1. getEnumValues(schema) — unwrap Zod wrappers (default, optional, nullable, pipe) and extract enum values
2. Integrate into getCommandInfo() — add enumValues field to OptionInfo and ArgumentInfo
3. Auto-display in help and docs — enum values appear after description in --help, LLM docs, and markdown reference
4. Auto-complete — options and arguments with z.enum() get autocomplete for free via bin/clever.js
5. Migrate drain create — use z.enum() instead of manual validation, leveraging the new framework

This replaces 7 manual enum value lists in descriptions with zero duplication, and provides better UX (help text, autocomplete) without extra code.

How to test

# Help text now shows enum values:
node bin/clever.js logs --help              # Format option shows (human, json, json-stream)
node bin/clever.js drain create --help     # drain-type shows all types
node bin/clever.js deploy --help           # exit-on and same-commit-policy show values

# Validation via Zod (not manual checks):
node bin/clever.js drain create invalid-type http://example.com
# Shows: Invalid option: expected one of "datadog"|"elasticsearch"|...

# Autocomplete works:
node bin/clever.js logs --format <TAB>     # completes with format values

[github-actions]

🔎 The preview has been automatically deleted.

[pdesoyres-cc]

Perfect. Well done!