feat(experiments): manage draft and running lifecycle

[joshdholtz]

Why

Let humans and agents create, duplicate, and manage experiments with clear review before enrollment changes.

Second PR in the stack: #179 → #180 → #181.

Changes

• Add create, duplicate, update, delete, start, pause, resume, and stop. Duplication composes get and create and always creates a draft.
• Accept simple create flags or a JSON config; rc schema experiments create|update describes config fields, placements, duration, conditions, and enums.
• Preview configuration and require confirmation or --yes before start, pause, resume, running updates, stop, and delete.
• Validate enrollment changes locally and add API, CLI, snapshot, and mock-server coverage.
• Reject deletion of non-draft experiments before confirmation and sanitize conflict names in human notices.

Review notes

• The development OpenAPI contract and backend validator allow enrollment from 1–100%. The dashboard guide says 10% minimum; the CLI follows the API contract.
• Create, draft update, and draft delete were exercised on a disposable project. Start reached the API but required a published control paywall; publishing was blocked by the separate agent permission layer, so start/pause/resume/stop and duplicate still need a live run.

Verification

go test ./..., go vet ./..., make gen, and make preview passed locally after rebasing onto current main.

---

Note

Medium Risk
Changes affect live customer enrollment and offering assignment when experiments are started, paused, resumed, or updated; guards and --yes gates reduce accidental mutation but the operational impact is still significant.

Overview
Adds full experiment lifecycle management to rc experiments: create, duplicate, update, delete, start, pause, resume, and stop, wired to new development v2 API routes and documented in the command surface, CLI coverage map, and beta OpenAPI overlay.

Create/update accept simple flags or a JSON --config (stdin/file); rc schema experiments create|update exposes config_fields for agents (metrics, placements, targeting, duration, etc.). Duplicate copies configuration into a new draft via get+create (no results/status). Local validation covers enrollment 1–100%, mutual exclusion of audience_id vs targeting_conditions, and placements rules (fallback offering IDs require at least one placement_offerings entry).

Lifecycle mutations show the current experiment configuration first and require confirmation or --yes (start, pause, resume, running enrollment updates, stop, delete). Updates are status-gated: drafts accept partial JSON; running experiments only enrollment_percentage; paused/stopped cannot be edited; delete is draft-only. The API client gains create/update/delete and action endpoints; human output tweaks include a consolidated conflicts line and sanitized notices/confirmations. Broad API, CLI, schema, and snapshot test coverage accompanies the change.

^{Reviewed by Cursor Bugbot for commit b772a60. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 4550e5a. Configure here.}

[JZDesign]

Sorry, My null parameter comment caused you to mark this one omitable/nullable too, but this one returns a 422 when omitted