feat: accept Pydantic models as alternatives to dicts in resource client methods

[vdusek]

Summary

• Resource client methods that accept dict input parameters now also accept the corresponding Pydantic model from _models.py as a union type (dict | Model)
• This gives users IDE hints about expected fields without requiring them to import models
• When a dict is passed, it is validated through the model; when a model is passed, it is serialized to dict for the API call
• encode_webhook_list_to_base64 updated to accept WebhookCreate models and None directly (with @overload for proper typing)

Changed parameters

File	Method	Parameter	New type
task.py	update, start, call, update_input	task_input	dict | TaskInput
task.py	start, call	webhooks	list[dict | WebhookCreate]
task_collection.py	create	task_input	dict | TaskInput
request_queue.py	add_request	request	dict | RequestDraft
request_queue.py	update_request	request	dict | Request
request_queue.py	batch_add_requests	requests	list[dict | RequestDraft]
actor.py	start, call	webhooks	list[dict | WebhookCreate]

Not changed (no matching model in _models.py)

• schema: dict in DatasetCollectionClient / KeyValueStoreCollectionClient — JSON Schema, no model
• batch_delete_requests(requests: list[dict]) — delete only needs IDs, no matching input model

Test plan

• All 234 unit tests pass
• Lint, type-check, async docstring sync all pass
• Verify IDE autocomplete shows model fields when hovering over union types

🤖 Generated with Claude Code