Detect missing shell-init during Hive setup/update

[notdp]

Problem

Codex sessions launched directly inside tmux can be non-native / not daemon-backed if the user's shell has not loaded Hive's shell integration:

eval "$(hive shell-init zsh)"

When this is missing, commands such as hive team fail with the degraded-runtime message and ask the user to add the shell-init snippet manually, then restart via hive codex resume.

This is correct as a recovery path, but the user only learns about it after hitting the degraded state.

Desired behavior

Hive should detect this during install/update or the nearest practical post-install/update flow, and proactively tell the user what to add.

Candidate surfaces:

• hive doctor: report whether the current shell startup file appears to load hive shell-init.
• hive shell-init: document the exact persistent install command more prominently.
• hive codex / degraded gate: keep the existing recovery message, but also point to the doctor/install check.
• install/update docs or plugin refresh flow: include a post-update check/reminder.

Caveat: pipx install / pipx upgrade may not provide a reliable hook to edit or inspect the user's shell startup files. If so, implement this as a Hive CLI / doctor / docs check rather than pretending install can always modify shell config.

Acceptance criteria

• A user who installs or updates Hive has a clear path to discover whether eval "$(hive shell-init zsh)" or the equivalent for their shell is configured.
• The check does not silently edit shell startup files without explicit user action.
• The degraded Codex error remains actionable and points to the same canonical setup path.
• zsh is covered first; fish/bash can be added if supported by hive shell-init.