fix(api): preserve evaluation quality in scoped entity summaries

[jonathanhaaswriter]

Summary

• fetch entity-summary graphs through the scoped entity graph path instead of the already-truncated current graph
• keep evaluation-quality signals available in scoped entity summaries
• add a regression test covering evaluation-quality preservation in scoped summaries

Testing

• go test ./internal/api -run "TestGraphIntelligenceEntitySummaryPreservesEvaluationQualityWhenScoped|TestGraphIntelligenceEntitySummaryUsesServiceInterface|TestPlatformEntitySummary" -count=1
• python3 scripts/devex.py run --mode changed --files internal/api/server_handlers_graph_intelligence.go internal/api/server_handlers_graph_intelligence_test.go internal/api/server_services_graph_intelligence.go

Closes #161

[jonathanhaaswriter]

I don't think the core prod path is fixed yet. The handler now calls graphIntelligenceEntityGraph(...), but the real service implementation still delegates CurrentEntityGraph(...) to CurrentGraph(...), so production still uses the scoped graph path instead of a fuller entity-aware graph source. That makes it look like the new behavior is only exercised by the test hook.

Can we wire CurrentEntityGraph to an actual full-graph/store-backed implementation before landing?