XY-840: [ELF benchmark vNext] Define real-world agent memory benchmark contract and scoring

[yvette-carlisle]

Summary:

• Adds the normative real_world_job benchmark contract with schema fields, suites, scoring dimensions, typed report states, and claim rules.
• Adds the operator-facing benchmark overview and links the new future-work contract from the benchmark indexes, README, live baseline guide, and adoption gate report without changing existing verdicts.

Validation:

• cargo make fmt
• cargo make lint-fix
• cargo make checks