Keep hard selection documentation focused on published rule

Explain two-stage hard-set selection

Cap hard-set median runtime at solve budget

Align hard split IDs with published ranking

Tighten hard split methodology wording

Fix hard split reproduction from HF snapshot

Document and reproduce frozen hard-50 selection

Add consolidated Gurobi reference tables

Sync remaining audited dataset files

Add wang2018 as the 180th canonical task

Move paper metadata under metadata directory