XY-927: [ELF benchmark P1] Add Letta-style core-vs-archival memory comparison

[yvette-carlisle]

Summary

• Add and preserve core_archival_memory benchmark coverage for core block attachment, scope, provenance, stale-core detection, archival fallback, and project-decision recovery.
• Keep Letta comparison rows blocked or not_tested until a contained export/readback path exists.
• Harden external adapter scenario validation so blocked scenarios require blocked outcomes and explicit outcome/position contradictions are rejected.
• Align benchmark reports, research guidance, and the benchmark spec with the manifest-backed evidence boundaries.

Verification

• cargo test -p elf-eval --test real_world_job_benchmark --all-features
• cargo make real-world-memory
• cargo make fmt
• cargo make lint-fix
• cargo make checks
• Independent fresh-context review checkpoint: clean for 05232fb