Backend: Referral fraud / sybil scoring and throttling

[joelpeace48-cell]

Context

/campaigns/:id/referrals tracks referrals but there is no fraud scoring. Referral programs attract
sybil/self-referral abuse that drains rewards.

Scope

• Heuristic risk score per referral (velocity, shared device/IP fingerprint, circular graphs, new-account bursts).
• Throttle/hold high-risk referrals for review; feed confirmed fraud into the contract clawback path.
• Configurable thresholds + admin review queue.

Acceptance criteria

• Obvious sybil patterns (rings, self-referral, IP bursts) are flagged and held, not auto-rewarded.

Verification

• Tests with crafted referral graphs assert correct scoring and hold/release behaviour.

[zachyo]

Hello, I've reviewed the issue and I can confidently submit a pr that meets the requirements. Please assign

[grantfox-oss]

🦊 GrantFox — @zachyo has been assigned to this issue as part of the Official Campaign campaign!

Next steps:

1. Open a Pull Request referencing this issue (e.g., Closes #764)
2. Your PR will be reviewed by the FinesseStudioLab maintainers

Good luck! Track your progress on GrantFox.

[joelpeace48-cell]

@zachyoyou are assigned, ensure all ci pipeline passes