proposal(starknet): strengthen deterministic deploy salt mixing

[omarespejel]

Proposal: improve deterministic deploy salt mixing for Starknet token deployment

This is a proposal issue for a small hardening enhancement.

Context

Current deterministic deployment salt uses only one portion of the token hash before conversion.

File context:

• starknet/src/omni_bridge.cairo (deploy salt derivation in deploy_token)

Proposal goal

Use a felt-compatible mix of both hash parts (or equivalent deterministic mixing) to improve entropy characteristics while preserving deterministic behavior.

Candidate approach

• derive salt from mixed low/high parts of token hash (deterministic, stable)
• keep output deterministic across environments and test fixtures

Non-goals

• no behavior change to token identity rules
• no change to signature/auth flow

Test expectations

• deterministic deployment remains stable for same token id
• collision-oriented regression tests for representative vectors

Maintainer input requested

1. Preferred mixing function for felt-compatible salt derivation?
2. Any compatibility constraints with existing deployed addresses on feat/starknet rollout?

Related Starknet epic: #554