fix(eval): pass simulation instructions to trajectory evaluator

[Chibionos]

Summary

• pass LLMMockingStrategy.prompt through AgentExecution.simulation_instructions when running evaluators
• add a runtime regression test for simulation instruction propagation
• add a trajectory evaluator test proving all built-in prompt placeholders are interpolated before the LLM call

Context

Slack thread: https://uipath-product.slack.com/archives/C08D98KT51U/p1781811738866539

Default Trajectory Evaluator reports missing prompt context for URT eval runs. Legacy eval-set migration stores simulationInstructions in EvaluationItem.mocking_strategy.prompt, but UiPathEvalRuntime.run_evaluator was not passing that value into AgentExecution, so {{SimulationInstructions}} could not be populated for trajectory prompts.

Validation

• uv run pytest tests/evaluators/test_evaluator_methods.py::TestLlmJudgeTrajectoryEvaluator tests/cli/eval/test_eval_tracing_integration.py::TestEvaluatorSpanCreation
• uv run ruff check src/uipath/eval/runtime/runtime.py tests/cli/eval/test_eval_tracing_integration.py tests/evaluators/test_evaluator_methods.py
• uv run ruff format --check src/uipath/eval/runtime/runtime.py tests/cli/eval/test_eval_tracing_integration.py tests/evaluators/test_evaluator_methods.py

Jira

Jira creation was blocked in the local workflow because no Jira creation provider is configured in this Codex environment. The local case has a ready-to-create payload.