Summary

Closes #72

Adds agent-friendly command variants with minimal, machine-readable output optimized for AI coding agents. This reduces context token usage and makes it easier to identify actual failures.

New Commands

Key Changes

Test Output Silencing:

• Added runSilent() helper using oclif's captureOutput to suppress stdout/stderr from command execution
• Updated stubParse helpers to set 'log-level': 'silent' by default
• Added global test setup (test/setup.ts) to set SFCC_LOG_LEVEL=silent
• Updated non-JSON mode tests to use runSilent() for clean output

Logger Enhancements:

• Added destination option to createLogger() for custom output streams
• Added createNullStream() and CapturingStream test helpers for logger testing
• Added resetLogger() export to reinitialize logger after config changes

Documentation:

• Updated AGENTS.md with new "Commands for Coding Agents" section
• Updated testing skill (.claude/skills/testing/SKILL.md) with output silencing patterns and test helpers reference

Cleanup:

• Removed unused operations/sites module (dead code with unimplemented stubs)

Example Output

Before (verbose):

  WebDavClient
    ✓ creates directory
    ✓ uploads file
    ✓ downloads file
  ... (hundreds of lines)
  
  728 passing (3s)

After (agent-friendly):

  728 passing (3s)

Failure output remains clear:

  728 passing (3s)
  2 failing

  1) test name:
     AssertionError: expected 2 to equal 3
     at Context.<anonymous> (test/file.ts:10:5)