fix(backups): accept pending lifecycle response values

[HarshMN2345]

Pending backup responses and restoration options currently fail model validation after successful API requests. Regenerate the two backup models from the corrected Cloud OpenAPI declarations in cloud#5760: pending size/start times are nullable and restoration options are an object. Keep policyId as the established string response, including "" for ad-hoc backups. Cloud's required/default response rule already converts stored null policy IDs to that empty string before serialization.

Extend the existing BackupsServiceTest cases for pending archive/restoration creation, completed ad-hoc archive retrieval, and populated restoration get/list responses. No new test class or runtime adapter is added.

Addresses MCP-Y, MCP-1K, MCP-1M, and MCP-1R.

Validation:

• Four regression cases fail with the original models and pass with the regenerated models; all 1,055 unit/service tests pass. Black passes for the changed files.
• Both model files are byte-identical to output from Cloud's OpenAPI formatter and locked Python generator.
• Actual Cloud Response::output() JSON passes through this SDK and the public MCP handler using a disposable local HTTP server, preserving null lifecycle fields, empty objects, and nested Unicode resource mappings. Completed ad-hoc archive responses also remain accepted by the released SDK.

Deployment order: Cloud#5760 must deploy first so empty options arrive as {}. Then publish this SDK under a new version and update/deploy MCP with that version in its lockfile. The currently locked SDK0.6.0 is not updated by merging these PRs alone.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; no outstanding correctness, security, or repository-rule violations remain.

Summary

• Makes archive size and start time nullable.
• Makes restoration start time nullable and models restoration options as an object.
• Adds observable service-level regression coverage for pending, completed, empty, and populated responses.
• Preserves the established serialized policyId: "" contract for ad-hoc backups.

_{Reviews (3) · Last reviewed commit: "fix(backups): preserve string policy IDs..."}

[HarshMN2345]

@greptile-apps review

Please re-review the current commit with the serialization evidence in the policy-ID discussion. Cloud's required/default response rule turns stored null policy IDs into empty strings; the corrected Cloud commit preserves that existing HTTP contract. The current SDK models are byte-identical to regeneration from that schema, and the old/new Cloud+SDK drill confirms released clients still accept completed ad-hoc archives. The four actual Sentry size/start/options regressions are still fixed.