Replace or retire legacy role-usage and eligibility-backed dataset queries

[gankoji]

Summary

Remove the remaining public dataset queries that still rely on legacy role-usage and legacy eligibility data, either by rebuilding them from Sprocket-native data or explicitly retiring them.

Why

A few public dataset queries still depend on legacy-only concepts or temporary legacy fallbacks, which will block final deprecation of MLEDB.

Candidate queries

• queries/public/role_usages.sql
• the legacy eligibility fallback inside queries/public/players.sql

Scope

• Decide whether each legacy-backed query should be rebuilt or retired.
• Replace legacy eligibility sourcing with Sprocket-native eligibility data once the source is trustworthy.
• Rebuild role-usage output from Sprocket-native data if that concept still matters.

Acceptance criteria

• No remaining public query requires mledb.team_role_usage or legacy mledb.eligibility_data.
• Retired queries are explicitly documented as retired.
• Dataset generation succeeds without these legacy dependencies.