FEAT GUI: Expose configured objective-scorer presets

Start with pyrit/registry/components/scorer_registry.py, pyrit/setup/initializers/scorers.py, and Scenario._get_default_objective_scorer(); trace how configured instances, presets, tags, and result families are represented. Review the existing /scores routes and model/Python/scorer/test instructions, then add fake-registered-scorer tests covering empty or incompatible registries, refusal polarity, credential safety, and the absence of scoring or model calls.

Is your feature request related to a problem? Please describe.

The custom-scenario flow needs a simple scorer choice. PyRIT already has ScorerRegistry and ScorerInitializer presets, but the GUI currently has no read API for choosing configured objective-scorer instances.

Work item 12 of 17 in #2744. This prepares the scenario milestone and remains not ready yet until promoted after item 11.

Describe the solution you'd like

Expose a small typed API for initialized, compatible objective-scorer presets from ScorerRegistry.instances.

• Return stable registry names, useful descriptions, scorer/result family, relevant tags, compatibility/prerequisite information, and the compatible configured default_objective_scorer selection.
• Offer ready instances, not just buildable class names. Listing must not instantiate scorers, invoke targets, or manufacture credential-backed configuration.
• Reuse existing presets/composites where available, such as inverted_refusal, task-achieved variants, acs_with_refusal, and scale_and_refusal. Availability depends on initialized targets and ScorerInitializer.
• Preserve objective-result meaning. Raw refusal=true means refusal and must not be exposed as attack success; use an appropriate existing inverter/composite or clearly validated objective preset.
• Report an empty/unusable registry clearly so the GUI can guide setup. Do not silently choose an incompatible default or expose secrets through identifiers/configuration metadata.

Acceptance criteria:

• The API distinguishes configured instances from merely discoverable classes.
• Compatible default tags and non-default presets are represented correctly.
• Missing dependencies, empty registries, incompatible result families, and refusal polarity are covered.
• The response contains no credential values and listing performs no scoring/model calls.
• Tests use fake registered scorers and require no paid services.

Describe alternatives you've considered, if relevant

Do not create a hardcoded frontend scorer list, a general scorer builder, a new scorer registry, or an unscored scenario mode. One selected objective scorer may already be a composite; arbitrary multi-scorer aggregation is out of scope.

Additional context

Start with pyrit\registry\components\scorer_registry.py, pyrit\setup\initializers\scorers.py, and Scenario._get_default_objective_scorer(). The existing /scores routes concern score operations, not a configured-scorer catalog. Follow model/Python/scorer/test instructions and keep registry responsibility separate from service DTOs.