Modularize live baseline benchmark script

[yvette-carlisle]

Summary

• Split the 3929-line live baseline benchmark script into focused shell modules under scripts/live-baseline/.
• Kept scripts/live-baseline-benchmark.sh as the orchestration entrypoint and extracted shared core, corpus, report, runner, OpenMemory probe, and per-project adapters.
• Updated source-inspection tests to assert against the new module locations without weakening the checked strings.

Batch size

• 15 files changed, 3878 insertions, 3866 deletions, 7744 total changed lines.
• Entry script reduced from 3929 lines to 81 lines.

Validation

• bash -n scripts/live-baseline-benchmark.sh scripts/live-baseline/*.sh scripts/live-baseline/projects/*.sh
• RUSTC=/run/current-system/sw/bin/rustc cargo nextest run -p elf-eval --test real_world_job_benchmark openmemory_ui_export_probe_has_dedicated_docker_task
• RUSTC=/run/current-system/sw/bin/rustc cargo nextest run -p elf-eval --test real_world_job_benchmark mem0_delete_audit_probe_requires_explicit_delete_history_event
• RUSTC=/run/current-system/sw/bin/rustc cargo make fmt-rust
• RUSTC=/run/current-system/sw/bin/rustc cargo make check (403 passed, 92 skipped)

Note

• A host smoke with ELF_BASELINE_PROJECTS=__none was attempted, but this macOS host lacks GNU timeout; the original script had the same required-command guard, so this is not introduced by the split.