feat(cli): add stack restart

[jgoux]

Add supabase stack restart for an existing managed stack, selected by the current project, --stack, or --stack-id. It stops and starts the same stack using its saved effective configuration, preserving its identity, data, service exclusions, activation settings, and preparation policy.

Restart does not reload project configuration or explicitly prepare artifacts. Runtime startup reuses cached artifacts and may fetch missing ones according to the saved policy. A normal start still loads project configuration and applies only its current flags, so start --exclude studio followed by stop and a plain start restores Studio when enabled in project configuration.

Stacks without a saved configuration are directed to stack start before they can be restarted.

Reimplemented directly on develop using the current stack command infrastructure.

Supersedes #6512

[github-actions]

🤖 AI Review

All seven source findings were verified and deduplicated into six entries: four confirmed and two refuted. The confirmed issues are two minor behavioral/error-handling concerns and two coverage/help-text nits; no critical or major defect was verified.

Findings

Severity	Location	Category	Sources	Claim
🟡 MINOR	apps/cli/src/commands/experimental/stack/restart/restart.handler.ts:146	behavior	claude	Restart can re-enable capabilities previously disabled through one-off stack start --exclude flags without warning.
🟡 MINOR	apps/cli/src/commands/experimental/stack/restart/restart.handler.ts:75	error-handling	claude+codex	Known StackRuntimeError and StackCleanupError failures lose the actionable recovery guidance provided by sibling lifecycle commands.
⚪ NIT	apps/cli/src/commands/experimental/stack/restart/restart.command.ts:8	help-text	claude	The --stack flag description does not state that omitting it targets the current project's stack.
⚪ NIT	apps/cli/src/commands/experimental/stack/restart/restart.integration.test.ts:177	test-coverage	claude	The restart handler suite lacks direct coverage for conflicting target flags, legacy output-flag rejection, and an inspectStack failure on the ID-target path.

Refuted findings (kept for transparency, not posted as review comments)

• apps/cli/src/commands/experimental/stack/restart/restart.handler.ts:149 (consistency): Clearing the successful restart task instead of completing it is inconsistent with stack lifecycle output conventions.
  Refuted: The surrounding stack commands establish both output patterns: stop.handler.ts:141 and 181 and destroy.handler.ts:131 clear their tasks before emitting dedicated text results. Restart likewise immediately renders a full status block, so clearing avoids a redundant completion line and is consistent with those siblings.
• apps/cli/docs/stack-commands.md:155 (documentation): The restart documentation is placed under an unrelated "Service selection and shutdown" heading.
  Refuted: The paragraph explicitly explains selecting a stack, stopping it, starting it again, and how service-selection flags are handled. Those subjects directly match the section's "Service selection and shutdown" heading.

Stats

Claude findings: 6 · Codex findings: 1 · Confirmed: 4 · Refuted: 2 · Uncertain: 0

---

Models: claude-opus-5 + gpt-5.6-sol · Trigger: auto · Workflow run

This review runs once per PR. A maintainer can request another with a /ai-review comment.