Add task hub and instance ID prefix orchestration query filters

從 OrchestrationQuery 和 build_query_instances_req() 開始，然後追蹤它們的同步和非同步用戶端呼叫方。將現有的查詢欄位與 .NET OrchestrationQuery 參考進行比較，並為每個篩選器分別使用以及同時使用時新增測試；當兩個欄位都序列化到 InstanceQuery 中，且省略的欄位維持目前的 requests 不變時，即視為完成。

Summary

The protobuf InstanceQuery supports taskHubNames and instanceIdPrefix, and the .NET Durable Task client exposes both filters. Python's OrchestrationQuery and request builder currently omit them.

Current behavior

Python orchestration queries support creation-time bounds, runtime status, result limits, and payload inclusion. There is no way to:

• restrict results to one or more task hubs;
• restrict results to instance IDs beginning with a prefix.

The generated protobuf already contains both fields, so no protocol update is required.

Expected parity

Extend OrchestrationQuery with Python-style equivalents of:

• task_hub_names: list[str] | None
• instance_id_prefix: str | None

Serialize these through build_query_instances_req() to InstanceQuery.taskHubNames and InstanceQuery.instanceIdPrefix for both sync and async clients.

.NET references:

• OrchestrationQuery
• GrpcDurableTaskClient.GetAllInstancesAsync

Acceptance criteria

• Callers can filter orchestration queries by task hub names.
• Callers can filter orchestration queries by instance ID prefix.
• Both fields are serialized into the existing protobuf request.
• Existing queries remain backward compatible when the fields are omitted.
• Unit tests cover each filter independently and together.