fix(console): fail fast in select_with_arrows when stdin is not a TTY

[yunaremaia]

Fixes #4152

Problem

specify init can hang indefinitely at arrow-key selection prompts when stdin is not attached to a TTY (agent harness, CI, piped input). select_with_arrows blocks in readkey() waiting for keypresses that never arrive — no timeout, no error, no output. This matters because agent-driven workflows are precisely the environments with no TTY, and Spec Kit is a natural fit for them.

Fix (defense in depth)

select_with_arrows now detects non-interactive stdin (sys.stdin.isatty()) and:

1. Resolves to default_key immediately when one is provided — making a fully scripted init expressible without needing a flag for every prompt (issue option 2).
2. Raises ValueError naming the missing interactive session when no default exists — fail fast with an actionable message instead of hanging (issue option 1).

Both call sites in init.py already guard with _stdin_is_interactive(), so this change cannot alter existing behavior for guarded paths — it closes the hang for any future or unguarded caller of the selector (issue: "there is no flag that reaches every prompt").

Tests

New file tests/test_console_non_interactive.py (3 tests):

• no-TTY without default → raises ValueError ("not a terminal")
• no-TTY with default → resolves to the default (no hang)
• TTY path → interactive loop untouched (get_key mocked to Enter)

Verified: 8/8 console tests green. tests/integrations/test_cli.py failures are pre-existing in this environment (84 failures with and without this change — identical count), so no regressions introduced.

[yunaremaia]

Hi maintainers — the workflow runs on this PR show action_required (first-time contributor policy). The change is small and self-contained:

• select_with_arrows now checks sys.stdin.isatty(): with a default_key it resolves immediately (no hang), without one it raises a clear ValueError instead of blocking forever on readkey() (fixes specify init hangs on select_with_arrows prompts in non-interactive environments — no flag reaches them, --force does not bypass them #4152).
• 3 new tests in tests/test_console_non_interactive.py — 8/8 console tests green locally, no regressions (the pre-existing test_cli.py failures in my environment are identical with/without the change).

Could you approve the runs when you get a chance? Thanks!