Long-Running Test spec, plan, and tasks

[brooke-hamilton]

Description

This PR adds a feature specification and plan for modernizing the Radius Long-Running Test (LRT) system to improve developer experience and align with contemporary workflow design principles.

The spec was generated using Spec Kit, which is why the folder structure does not align with our existing folder structure.

Overview

The specification defines a modernized LRT system that enables:

• Developer self-service: Developers can provision dedicated AKS test environments, run tests locally, and tear down infrastructure using simple Make commands
• Fork-compatible workflows: Contributors can test workflow changes on forks without access to radius-project organization resources
• Automated release validation: System automatically detects version mismatches and upgrades clusters before running tests
• Improved observability: Comprehensive health checks, diagnostic artifacts, and actionable error messages

Key User Stories (Priority Order)

Developer Infrastructure Setup

• Provision dedicated AKS clusters with full isolation per developer
• Run complete LRT suite locally with simple commands (make lrt-deploy-infra, make lrt-run-all, make lrt-destroy-infra)
• Infrastructure deployment completes in 30-45 minutes with idempotent operations

Fork-Based Workflow Testing

• Execute LRT workflows on forks using only fork owner's resources
• Manual workflow triggering via workflow_dispatch on any branch
• Scheduled triggers ignored on forks to prevent unexpected resource usage

Run tests on the current release

• Compare CLI vs cluster Radius versions at workflow start
• Automatic cluster upgrade using rad upgrade when versions differ
• Configurable rollback on test failures

Local Test Execution

• Run individual test suites (make lrt-test-corerp) or full suite (make lrt-run-all)
• Real-time progress output and fast feedback loops
• Test execution completes within 120 minutes