feat: get plan bulk with cache

[tomerqodo]

Benchmark PR from qodo-benchmark#445

[augmentcode]

🤖 Augment PR Summary

Summary: Adds a cached bulk subscription-plan fetch path to reduce Billing API load in batch scenarios.

Changes:

• Adds Redis-backed caching for tenant subscription plans with a 10-minute TTL (`get_plan_bulk_with_cache`).
• Introduces a Redis key convention (`tenant_plan:{tenant_id}`) and helper to build cache keys.
• Adds per-tenant validation error handling when parsing Billing API bulk responses.
• Implements cache read via `mget` and cache writes via Redis pipeline + `setex`.
• Adds new testcontainers-based integration tests covering cache hit/miss, Redis failures, invalid cache data, and TTL expiry.
• Adds a unit test for handling invalid tenant plan payloads from the Billing API.

Technical Notes: Cached values are JSON-serialized plan dicts and re-validated via Pydantic TypeAdapter on read; Redis failures fall back to the Billing API.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. 2 suggestions posted.

Comment augment review to trigger a new review at any time.

[augmentcode]

get_plan_bulk() is annotated/documented to return SubscriptionPlan values, but on validation failure it stores None in results which can break callers that assume a dict with plan/expiration_date (and it also diverges from the new unit test expectation of “skip invalid tenant”). Consider making the invalid-tenant behavior consistent (either skip, or change typing/consumers to explicitly handle None).

Other Locations

• api/services/billing_service.py:366
• api/services/billing_service.py:376

_{🤖 Was this useful? React with 👍 or 👎}

[augmentcode]

This test expects the invalid tenant to be omitted ("tenant-invalid" not in result), but the current get_plan_bulk() implementation adds the key with a None value on validation failure, so this will fail. Consider aligning the test with the intended behavior (skip vs None).

_{🤖 Was this useful? React with 👍 or 👎}