Refactor the evaluation results handling and upload to Braintrust

[haoranpb]

Refactoring the evaluation result into a different module with some improvements, address some other concerns along the way:

• Better exception handling with Timeout

The eval results should be uploaded to:

• Braintrust with bceval

Results are summarized into evaluation_summary.json

[Copilot]

Pull Request Overview

This PR refactors the evaluation results handling by extracting the result-related functionality from bcbench.evaluate into a new bcbench.results module, and introduces a new CLI command group for result operations. The changes improve code organization and add support for exporting results to bceval format with Braintrust integration.

Key changes:

• Moved EvaluationResult and EvaluationResultSummary from evaluate/evaluation_result.py to results/evaluation_result.py
• Split display logic into dedicated results/display.py module
• Added results/result_writer.py for bceval format export
• Moved evaluate summarize command to new result summarize command in commands/result.py
• Updated config to dynamically load instance pattern from dataset schema for result file filtering
• Added Braintrust upload step to GitHub workflow with bceval integration

Reviewed Changes

Copilot reviewed 17 out of 18 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/bcbench/results/evaluation_result.py	New location for EvaluationResult and EvaluationResultSummary classes with from_json and from_results factory methods
src/bcbench/results/display.py	Extracted display logic (console and GitHub summaries) into dedicated module
src/bcbench/results/result_writer.py	New module for writing results in bceval format
src/bcbench/results/init.py	Public API exports for results module
src/bcbench/commands/result.py	New CLI command group with result summarize command
src/bcbench/config.py	Added dynamic instance_pattern loading from dataset schema and result_pattern to FilePatternConfig
tests/test_evaluation_summary.py	Updated tests to use new module paths and EvaluationResult objects instead of raw dicts
src/bcbench/evaluate/evaluation_pipeline.py	Updated imports and result file naming to use config pattern
docs/_data/results.json	Added leaderboard results data
evaluator/scores.py	New bceval evaluator for resolution rate metric
.github/workflows/*.yml	Updated workflows to use new result summarize command and added Braintrust integration