Materialize Letta core/archive export benchmark

[yvette-carlisle]

Summary

• add Docker-contained Letta core/archive export-readback smoke task and optional Letta compose profile
• publish the June 19 Letta materialization evidence report and JSON companion
• update external adapter manifest/tests so all six Letta core/archive scenarios are typed blocked until live export/readback source ids exist

Benchmark

• cargo make smoke-letta-core-archive-export-readback: 6 jobs, 0 pass, 0 wrong_result, 6 blocked, 14/14 evidence/source-ref/quote coverage
• cargo make real-world-memory-core-archival: 6 jobs, 6 pass, 0 blocked, 0 wrong_result, evidence coverage 1.0

Validation

• python -m py_compile scripts/letta-core-archive-export-readback-smoke.py
• cargo make smoke-letta-core-archive-export-readback
• jq empty apps/elf-eval/fixtures/report_snapshots/2026-06-19-letta-core-archive-export-readback-report.json && jq empty apps/elf-eval/fixtures/real_world_external_adapters/memory_projects_manifest.json
• cargo test -p elf-eval --test real_world_job_benchmark letta_core_archive_export_readback_report_preserves_blocked_gates -- --test-threads=1
• cargo test -p elf-eval --test real_world_job_benchmark real_world_report_includes_external_adapter_coverage_manifest -- --test-threads=1
• cargo test -p elf-eval --test real_world_job_benchmark external_adapter_run_summarizes_nonzero_scenario_losses -- --test-threads=1
• cargo test -p elf-eval --test real_world_job_benchmark external_adapter_manifest_rejects_unmeasured_win_loss_scenario_outcomes -- --test-threads=1
• cargo test -p elf-eval --test real_world_job_benchmark generated_json_report_renders_markdown -- --test-threads=1
• cargo make real-world-memory-core-archival
• cargo make fmt-check
• cargo make check-docs
• decodex docs check
• cargo make lint
• cargo make check