SubStringScorer identifiers omit text matcher configuration

[Garyouki]

Describe the bug

SubStringScorer._build_identifier() records only the text matcher's class name. The built-in matchers' case_sensitive, ignore_whitespace, threshold, and n settings are absent from both the component hash and evaluation hash.

Two scorers can therefore return opposite verdicts for the same input while sharing an evaluation identity. ScorerEvaluator._should_skip_evaluation() looks up existing metrics by this hash, so an evaluation can reuse metrics from a different matcher configuration when the remaining cache conditions match.

Steps/Code to Reproduce

import asyncio
from unittest.mock import MagicMock, patch
from pyrit.analytics import ExactTextMatching
from pyrit.memory import CentralMemory, MemoryInterface
from pyrit.score import SubStringScorer

async def reproduce_async():
    with patch.object(CentralMemory, "get_memory_instance", return_value=MagicMock(spec=MemoryInterface)):
        a = SubStringScorer(substring="Hello", text_matcher=ExactTextMatching(case_sensitive=True))
        b = SubStringScorer(substring="Hello", text_matcher=ExactTextMatching(case_sensitive=False))
        print((await a.score_text_async("hello"))[0].get_value())
        print((await b.score_text_async("hello"))[0].get_value())
        print(a.get_identifier().hash == b.get_identifier().hash)
        print(a.get_identifier().eval_hash == b.get_identifier().eval_hash)

asyncio.run(reproduce_async())

Expected Results

Verdicts are False and True; both hash comparisons are False because the behavioral configurations differ.

Actual Results

Verdicts are False and True, but both hash comparisons are True. Reproduced for all five built-in matcher settings listed above (case sensitivity for each matcher).

Versions

PyRIT main at b6a18a20, Python 3.12.3, Linux. No external model or service is needed.