Experimentation Form - Models and CRUD

[Zaimwa9]

Implement the skeleton models for experimentation

Description
Minimal Experiment model and CRUD endpoints in the experimentation app. Just enough to wire up the frontend wizard's Step 1 (Setup) and the experiments list page. Audience, metrics, and launch logic come in follow-up issues.

Model

class ExperimentStatus(TextChoices):
      CREATED = "created"
      RUNNING = "running"
      PAUSED = "paused"
      COMPLETED = "completed"

class Experiment:
      environment: FK → Environment
      feature: FK → Feature              # only multivariate flags
      name: CharField(255)              # required
      hypothesis: TextField               # required
      status: CharField (default=CREATED)
      created_at: DateTimeField(auto_now_add)
      updated_at: DateTimeField(auto_now)
      started_at: DateTimeField(null)
      ended_at: DateTimeField(null)

• FK to Feature a flag can have past completed experiments, but only one active (created/running/paused) per flag+environment.
• UniqueConstraint: (feature, environment) where status NOT IN (completed).

Endpoints — nested under environment like warehouse connections:

  /environments/<env_key>/experiments/
  - GET — list (filterable by status)
  - POST — create
  - GET /:id — retrieve
  - PATCH /:id — update (name, hypothesis, status)
  - DELETE /:id — soft delete

Not in scope: audience conditions, sample %, variation weights, metrics, launch validation.