MPT-21087 – Reduce cloud API rate-limit failures and improve their visibility

[ffaraone]

Description

Problem

When a tenant has multiple cloud accounts (e.g. several Azure subscriptions under the same tenant), their billing imports run in parallel and share the same cloud-provider API rate-limit quota. Under load, this causes cascading HTTP 429 errors that exhaust all retry attempts, leaving imports in a failed state and surfacing an ERROR event in the organisation's event log, even though the failure is transient and automatically retried.

What changed

• Parallel imports per tenant are now bounded. Accounts that share a rate-limit quota (identified by a common parent account) are imported with a configurable maximum concurrency instead of all at once. This significantly reduces the chance of hitting provider rate limits in the first place. The limit is tunable via the /diworker/max_tenant_import_workers configuration key (default: 1).
• Rate-limit failures are no longer reported as errors. When an import fails solely because the cloud provider throttled the requests, the event log now records a WARNING ("was throttled by the cloud provider") rather than an ERROR ("failed"). Genuine billing API failures are unchanged and still appear as errors.

Related issue number

Special notes

Checklist

• The pull request title is a good summary of the changes
• Unit tests for the changes exist
• New and existing unit tests pass locally