test: fee rounding proptest (#746)

[karanjadavi]

Summary

Closes #746. Adds a property-based test suite for FeeCalculator that asserts platform fees never exceed the configured basis-point rate.

What changed

• Replaced the previously-registered but non-compiling contracts/predictify-hybrid/src/tests/fee_calculator_proptest.rs (it referenced proptest APIs that don't exist: proptest::strategy::StrategyWrapper, .proptest_individuals(), testutils::test()) with a working suite built on the real proptest! macro, following the same pattern already used by fee_calculator_proptest.rs's sibling fees.rs unit tests and ValidationTestingUtils::create_test_market.

Why not contracts/predictify-hybrid/tests/proptest_fee_rounding.rs as suggested in the issue

fees, types, and markets are declared as private mod (not pub mod) in lib.rs, so an external integration test under tests/ cannot reach FeeCalculator at all. Every existing fee-related test (fee_idempotency_tests.rs, fee_config_commit_reveal_tests.rs) lives under src/tests/ for the same reason. I kept that pattern instead of introducing a file that can't compile.

Properties covered

1. fee <= total_staked for every generated stake
2. fee never exceeds the configured bps rate (fee * 10_000 <= total_staked * PLATFORM_FEE_PERCENTAGE)
3. Rounding is strictly floor (fee is the largest integer satisfying the bound above)
4. platform_fee + user_payout_amount == total_staked (breakdown reconciles exactly)
5. Fee is monotonically non-decreasing as stake increases
6. Focused edge cases: zero stake, i128::MAX (overflow guard), MIN_FEE_AMOUNT boundary, small-pool rounding

Verification

This branch is based on current master (43fdfdd), which does not compile as-is (325 pre-existing error[...] diagnostics from unrelated modules — missing mod declarations for disputes, edge_cases, market_analytics, performance_benchmarks, graceful_degradation, plus renamed/removed functions like require_primary_admin). I confirmed this new test file introduces zero new error locations by diffing cargo test -p predictify-hybrid --no-run output with and without this change (identical --> locations both ways). Once the unrelated build breakage upstream is fixed, this suite should compile and run cleanly — happy to re-verify at that point if useful.

[drips-wave]

@karanjadavi Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[karanjadavi]

Heads up: CI will likely fail on this PR — master (43fdfdd) currently has 325 pre-existing compile errors unrelated to this change (missing mod declarations for disputes/edge_cases/market_analytics/performance_benchmarks/graceful_degradation, plus several renamed/removed functions like require_primary_admin). Verified this file adds zero new error locations by diffing cargo build output with/without it. Flagging in case a separate issue should track the master breakage.

[karanjadavi]

Update: CI is failing for an additional, separate reason beyond the local mod/function issues I flagged earlier — the ethnum v1.5.0 dependency (pulled in transitively via num-bigint) fails to compile on the CI runner with error[E0512] (transmute size mismatch in TryFromIntError). This is unrelated to this PR's change as well; it's a dependency/rustc-version compatibility issue in the workspace's Cargo.lock, not something introduced here. Not attempting to fix it in this PR since pinning a different ethnum version is a separate, workspace-wide change outside this task's scope.