Evaluation engine criteria causes errors when intermediate evaluation_steps generation returns poorly formatted JSON/dict

[griptapeOsipa]

Description:

When "criteria" is used in the evaluation engine, during the automagic generation of evaluation_steps, the intermediate JSON/dict is sometimes improperly formatted, leading to runtime errors.

Steps to Reproduce:
Running this as-is will work (sending something to the evaluation steps directly). Commenting "evaluation_steps" out, and "criteria" in, shows the problem:

from griptape.structures import Pipeline
from griptape.engines import EvalEngine
from griptape.tasks import PromptTask
from griptape.rules import Rule

from dotenv import load_dotenv
load_dotenv() # Load the environment variables

rules    =  [
                "Answer with a json object, with no additional markup.",
                "Talk like a pirate.",
            ]

pipeline = Pipeline(
    tasks = [
        PromptTask(
            "Respond to this user: '{{ args[0] }}'"
            "{% if args[1] %}Use this feedback when answering.{{ args[1] }}{% endif %}"
        ),
    ],
    rules=[ Rule( rule ) for rule in rules ],
)

engine = EvalEngine(
    #criteria=[
    evaluation_steps=[
        f"Determine whether the following rules have been met: {rules}",
    ]
)

pipeline.run( "Hi there" )
score, reason = engine.evaluate(
    input         = pipeline.tasks[0].input.value,
    actual_output = pipeline.output.value,
)

Expected Behavior: The intermediate JSON/dict for evaluation_steps should always include a steps key, an be properly formatted for use.

Environment:

Griptape version: 1.1.1
Python version: 3.10-3.12
OS: OSX

Debug text pulled to look at the json directly

('{"type": "object", "properties": {"steps": {"type": "array", "items": '
 '{"type": "string"}}}, "required": ["steps"], "additionalProperties": false, '
 '"$id": "Output Format", "$schema": '
 '"http://json-schema.org/draft-07/schema#"}')

How the error reads:

Traceback (most recent call last):
  File ".../GitHub/griptape-TS-1_1/evalEngine.py", line 33, in <module>
    score, reason = engine.evaluate(
  File ".../GitHub/griptape-TS-1_1/.venv/lib/python3.10/site-packages/griptape/engines/eval/eval_engine.py", line 86, in evaluate
    self.evaluation_steps = self._generate_steps(evaluation_params)
  File ".../GitHub/griptape-TS-1_1/.venv/lib/python3.10/site-packages/griptape/engines/eval/eval_engine.py", line 115, in _generate_steps
    return parsed_result["steps"]
KeyError: 'steps'

[collindutter]

Fixed via #1519