Fix integration tests hanging on machines with a TTY

[khaong]

Summary

• After fix flaky test #404 moved detectOrSelectAgent to the command handler, integration tests calling entire enable --strategy now hit the interactive agent selection form when a real TTY is available, causing hangs
• Sets ENTIRE_TEST_TTY=0 in cliEnv() so all CLI subprocesses skip the TTY check by default
• Tests that need TTY simulation (hook tests) already override this per-subprocess with ENTIRE_TEST_TTY=1

Test plan

• mise run fmt && mise run lint passes
• mise run test:ci passes (unit + integration)
• Verified all ENTIRE_TEST_TTY=1 usages are on per-subprocess env, not via cliEnv()

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Test-only environment change that reduces flakiness by avoiding interactive prompts; low risk outside integration test behavior.

Overview
Prevents integration tests from hanging on machines with a real TTY by forcing non-interactive behavior for all CLI subprocesses.

TestEnv.cliEnv() now always sets ENTIRE_TEST_TTY=0, ensuring commands that would otherwise prompt (e.g., agent selection) take the no-TTY path; tests that need TTY behavior can still override this per subprocess.

^{Written by Cursor Bugbot for commit 8253d4b. Configure here.}

[Copilot]

Pull request overview

This PR fixes a regression introduced in PR #404 where integration tests would hang on machines with a real TTY. After PR #404 moved detectOrSelectAgent to the enable command handler, tests calling entire enable --strategy started hitting the interactive agent selection form when running on machines with a TTY available, causing the tests to block indefinitely.

Changes:

• Sets ENTIRE_TEST_TTY=0 in the default test environment to prevent interactive prompts from blocking tests
• Tests that specifically need TTY simulation (like hook tests) already override this per-subprocess with ENTIRE_TEST_TTY=1