test: Added missing unit and integration tests

[usanvict]

Summary

Adds comprehensive unit and end-to-end test coverage for the Apify LangChain tools, targeting gaps not covered by the existing test_tools.py and test_client.py suites.

New unit tests

• test_clamping.py -- Tests _clamp_depth (previously untested), exact at-boundary values for all clamp functions (_clamp_timeout, _clamp_memory, _clamp_items, _clamp_depth), and relaxed custom limit pass-through.
• test_input_validation.py -- Exhaustive Pydantic input schema validation: missing required fields, wrong types, offset boundaries, empty/long string IDs, unicode/special characters, dict structure edge cases (deeply nested, None values, large dicts), URL format boundaries, ApifyDatasetLoader constructor validation, and confirmation that schemas intentionally defer range-checking to clamping.
• test_integration_mock.py -- Tests the LangChain .invoke() error-handling path (handle_tool_error=True converts ToolException to a string instead of raising) and verifies output key sets match the documented API contract for RunActor, RunActorAndGetDataset, and GetDatasetItems.
• test_run_meta.py -- Edge cases for the _run_meta() utility: empty dict returns all Nones, extra API keys are filtered out, and non-SUCCEEDED statuses (FAILED, TIMED-OUT, RUNNING, ABORTING) are preserved.
• test_two_step_orchestration.py -- Tests the RunActor -> GetDatasetItems two-step orchestration pattern: dataset_id flows correctly between steps, failed Step 1 produces non-JSON error output, and each run produces a unique dataset ID.

New E2E integration tests

All E2E tests require APIFY_API_TOKEN and are automatically skipped when the token is not set.

• test_e2e_auth.py -- Missing token raises ValueError; invalid token returns a non-JSON error string.
• test_e2e_happy.py -- Happy-path coverage for all 6 core tools (RunActor, GetDatasetItems, RunActorAndGetDataset, ScrapeUrl, RunTask, RunTaskAndGetDataset) against the live Apify API using apify/python-example.
• test_e2e_negative.py -- Parametrized test verifying that invalid actor/dataset/task IDs return graceful error strings (no exception propagation).
• test_e2e_stability.py -- Repeated invocations produce structurally consistent output shapes and unique dataset IDs; fetching the same dataset multiple times returns identical items.
• test_e2e_timeout.py -- Boundary conditions against the live API: offset beyond dataset returns empty, limit larger than dataset returns available items, and explicit memory_mbytes=128 is accepted.