Assessment: Add support for Anthropic

[vprashrex]

Is your feature request related to a problem?
The assessment module's batch pipeline only supports OpenAI and Gemini providers, preventing assessments configured with Anthropic models from running as batch jobs. This limitation hinders cost-efficient bulk evaluations of Claude models.

Describe the solution you'd like

• Implement an AnthropicBatchProvider that meets the BatchProvider interface, using the Anthropic Message Batches API for batch creation, status polling, and result retrieval.
• Add build_anthropic_jsonl to format assessment items for the Anthropic batch request and route assessments to the new provider.
• Include map_kaapi_to_anthropic_params for translating Kaapi model parameters to Anthropic API format.
• Create build_anthropic_attachment_parts to convert assessment attachments into Anthropic content blocks.
• Extend output parsing to manage Anthropic batch responses, including error handling and text extraction.
• Provide unit tests for the new provider, mappers, and parsing in the relevant test directories.

Original issue

Describe the current behavior
The assessment module's batch pipeline (backend/app/core/batch/) only supports OpenAI and Gemini providers, so assessments configured with Anthropic models cannot run as batch jobs. This blocks cost-efficient bulk evaluation runs on Claude models and forces Anthropic-backed assessments through the synchronous path.

Describe the enhancement you'd like

• Add an AnthropicBatchProvider implementing the BatchProvider interface (backend/app/core/batch/anthropic.py), covering batch creation via the Anthropic Message Batches API, status polling, and result retrieval.
• Add build_anthropic_jsonl in backend/app/crud/assessment/batch.py to format assessment items into Anthropic batch request format, and extend submit_assessment_batch to route Anthropic-provider assessments to the new provider.
• Add map_kaapi_to_anthropic_params in backend/app/services/assessment/mappers.py to translate Kaapi model parameters (including max_tokens defaulting via DEFAULT_ANTHROPIC_MAX_TOKENS) to Anthropic API format.
• Add build_anthropic_attachment_parts in backend/app/services/assessment/utils/attachments.py to convert assessment attachments into Anthropic content blocks.
• Extend output parsing in backend/app/crud/assessment/processing.py to handle Anthropic batch responses, including error results and text extraction from content blocks.
• Cover the new provider, mappers, and parsing with unit tests (app/tests/core/batch/test_anthropic.py, app/tests/assessment/).