feat(contribute): add sharing.contributeHint.enabled to opt out of the share-learnings nudge

[Rererr]

Summary

Add sharing.contributeHint.enabled (team default), contributeHintEnabled (member override) and TEAMAI_CONTRIBUTE_HINT_DISABLED=1 (kill switch) so a team can turn off the post-session /teamai-share-learnings nudge without disabling the whole built-in Stop hook. Today the only way to silence the nudge is builtin.disabled: ["Hook dispatch stop"] in hooks/hooks.yaml, which also drops the CLI update check, votes sync and dashboard reporting.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature causing existing behavior to change)
• Documentation only
• Refactor / internal cleanup

Test Plan

• npx tsc --noEmit passes

• npx vitest run passes (full suite; the only failures on my macOS machine are lock-atomic.test.ts and push-*.test.ts concurrency/timing tests, which fail identically on a pristine main checkout and are unrelated to this change)

• Added/updated tests for the change

• contribute-hint-toggle.test.ts (new): resolution order (default true → team → member override → env kill switch) and schema parsing, including that existing teamai.yaml / config.yaml without the new fields still parse.

• hook-handlers.test.ts: contribute-check returns null and never calls contributeCheckForSession when the team turned the hint off; a member override re-enables it; config load failure keeps the old behavior (hint still emitted); TEAMAI_CONTRIBUTE_HINT_DISABLED=1 silences it; pending-hint consumes a stashed hint without delivering it while still delivering votes hints.

Mutation check: removing the guard in contributeCheckHandler makes two of the new tests fail.

Related Issues

None directly. Related context: #354 (share-learnings hint delivery on stdout-less tools) — this PR keeps that stash path intact and only decides whether the stashed hint is shown.

Notes for Reviewers

• Shape follows the existing two-tier pattern of sharing.recall / sharing.coAuthor (optional block, no .default so existing TeamaiConfig literals stay valid; resolved through isContributeHintEnabled()).
• The handler resolves the switch per run via autoDetectInit() (project scope first, then user), so flipping it in teamai.yaml takes effect on the next pull with no hook re-injection. If config cannot be loaded it falls back to enabled, so half-initialized installs behave exactly as before.
• On stdout-less tools (codebuddy / workbuddy) the pending-hint handler still consumes the stash when disabled, so a hint stashed before the switch flipped is never delivered later when it is turned back on. Votes hints on the same handler are unaffected.
• Scope: only the nudge. Friction scoring, teamai contribute --file, and manual /teamai-share-learnings keep working. Docs updated in both usage guides (new "Turning the hint off" subsection + config reference) and both READMEs (one sentence).
• Motivation: our team runs a personal retrospective loop that files ordinary PRs to the team repo, so the automatic nudge is noise for us — but we still want update checks and reporting from the Stop hook.