Add transactional reload status and rollback tracking to Prometheus

Normalized profile

What this task asks for

Add an opt-in transactional config reload mode with durable reload outcomes, rollback tracking, and an HTTP status endpoint. Add an opt-in transactional reload mode that executes reloaders in sequence and records a single outcome for the whole attempt. The most recent outcome must be observable via HTTP and durable across restarts so operators can diagnose failures after a restart. Persist the most recent reload outcome as JSON under the configured TSDB storage directory. The persisted JSON must include at least: last_reload_id, last_reload_successful, error_category (it is recommended to persist the same fields as the /api/v1/status/reload response). Error_category must be one of: none, load_error, apply_error, rollback_error. Missing or corrupted persisted state must not prevent startup or the endpoint from working. Enabling transactional-reload-config must be reflected in GET /api/v1/features as prometheus.transactional_reload_config.