feat(thinkfleet-memory): seed plans + entitlements via Rake task

[rrader26]

Summary

• Adds lib/tasks/thinkfleet_memory.rake with one idempotent task that provisions the four ThinkFleet Memory tiers, their billable metrics, features (+ privilege catalogs), entitlements, and a thinkfleet-memory billing entity for customer isolation.
• Designed to be run against any organization in this Lago org by id: bin/rails 'thinkfleet_memory:seed[<organization_id>]'. Append ,dry_run to preview without persisting.
• Idempotent: every create is find_or_create_by keyed by code. Entitlement values update in place — re-run anytime to bump quotas.

Why

Mirrors the EngageIt isolation pattern (billing entity per product) so Memory customers, subscriptions, and invoices live under thinkfleet-memory while plans share the org. The feature codes (memory.subjects, memory.predictions, etc.) are the wire contract with the memory-thinkfleet API consumer — the engine reads these at request time via /plans/<code>/entitlements to gate features and enforce limits.

What gets created

Type	Codes
Billing entity	thinkfleet-memory
Billable metrics	memory.events_ingested.used, memory.predictions.used (sum_agg on amount)
Features	memory.subjects, memory.active_triggers, memory.events_ingested, memory.predictions, memory.health_vertical, memory.finance_vertical, memory.on_device, memory.vpc_deployment, memory.audit_retention
Plans	memory-dev (free), memory-team ($199/mo), memory-business ($1,499/mo), memory-enterprise (custom)

Privilege values per plan match the spec at `docs/memory-plan-spec.md` in the memory-thinkfleet repo.

Test plan

• Pull the branch, run `bin/rails 'thinkfleet_memory:seed[<qa_org_id>,dry_run]'` against QA Lago — confirm the printed plan reflects intent
• Run without `dry_run` → confirm all four plans exist with correct entitlement values
• Re-run → confirm output shows only ✓ (no + / ~) — proves idempotency
• Confirm `BillingEntity.find_by(code: "thinkfleet-memory")` exists
• From the memory.thinkfleet.ai API: `GET /api/v1/plans/memory-dev/entitlements` returns the entitlements as expected

After merge

1. Deploy this branch to the Lago environment
2. From a deployed pod: `bin/rails 'thinkfleet_memory:seed[<your_org_id>]'`
3. The memory-thinkfleet consumer API reads these via the existing Lago entitlement checker — no consumer-side changes needed

🤖 Generated with Claude Code