Add eval validate and create commands

[HartBrook]

Summary

Improves developer experience for creating and maintaining behavioral evals:

• stag eval validate - Validate eval YAML files before running (saves API credits)

  • Checks assertion types, required fields, YAML structure, naming conventions
  • Provides helpful suggestions for common typos (e.g., llm_rubric → llm-rubric)
  • Distinguishes between errors (blocking) and warnings (non-blocking)

• stag eval create - Create new evals from templates with interactive wizard

  • Four built-in templates: security, quality, language, blank
  • --template flag to skip wizard and use template directly
  • --from flag to copy and customize existing evals
  • --project flag to save to .staghorn/evals/
  • --team flag to save to ./evals/ for team/community sharing

• Example evals - Added example/team-repo/evals/ with team eval patterns

Test plan

• go build ./... passes
• go test ./... passes (22 new tests for validation and templates)
• stag eval validate validates all evals correctly
• stag eval validate catches invalid assertion types with suggestions
• stag eval create --template security --name test --description "test" creates valid eval
• stag eval create --team saves to ./evals/
• Example evals in example/team-repo/evals/ are valid YAML