XY-846: [ELF benchmark suite] Add memory evolution and temporal staleness cases

[yvette-carlisle]

Summary

• Adds checked-in real-world memory evolution fixtures covering changed preferences, blocked-to-done issue state, superseded deployment guidance, overturned benchmark verdicts, and temporal relation limitations.
• Extends the real_world_job runner/report with memory evolution counters, job-level encoding status, follow-up rows, and temporal validity not_encoded reporting.
• Documents the new cargo make real-world-memory-evolution benchmark task and aligns suite not_encoded semantics.

Verification

• cargo make real-world-memory-evolution
• cargo nextest run -p elf-eval --test real_world_job_benchmark --all-features
• Semantic drift helper plus manual evidence mapping
• cargo make fmt
• cargo make lint-fix
• cargo make checks