fix(client): use exact fixed-point order arithmetic

[cesarenaldi]

Summary

• accept valid numeric and numeric-string prices after bounded snapping of insignificant IEEE-754 drift, while continuing to reject materially off-grid values
• validate price range and tick-grid membership with six-decimal fixed-point bigint arithmetic
• migrate limit and market order amount calculations to exact fixed-point arithmetic with explicit directed rounding
• add regression coverage for both order calculators, both sides, protection modes, and every supported tick size

Implementation plan

• Add behavior-preserving characterization coverage for current order amount calculations
• Introduce shared fixed-point primitives and migrate limit/market amount calculations
• Add bounded price snapping and tick-grid regression coverage

Verification

• pnpm test:client -- --run
• pnpm lint
• pnpm typecheck

Fee calculation and maxSpend resizing are intentionally deferred to stacked draft PR #306.

Fixes #297

Linear: DEV-576

---

Note

Medium Risk
Changes how order prices and maker/taker amounts are computed before submission; incorrect rounding could cause rejections or wrong fills, though behavior is heavily regression-tested.

Overview
Replaces float-based order price validation and amount encoding with six-decimal fixed-point bigint math so limit and market orders match the exchange tick grid without IEEE-754 rounding surprises.

Price handling: validatePriceOnTickGrid now normalizes via toScaledPrice / toScaledAmount, tolerates insignificant drift within MAX_PRICE_DRIFT, rejects excess precision with unsupported precision, and checks grid membership with integer modulo. It returns a ScaledPrice instead of a raw number. Limit/market prep keep scaled prices internally and only convert back to number where fee helpers still expect floats.

Amount encoding: Limit and market offered/requested amounts move into shared computeLimitOrderAmounts / computeMarketOrderAmounts in amounts.ts, using mulDiv and quantize with explicit round-down vs round-up (including protectPrice on market buys). Public size/amount inputs are still quantized to two decimals before calculation. The old math.ts helpers (parseAmount, roundDown, decimalPlaces, etc.) are removed.

Tests: New coverage for fixed-point primitives, all supported tick sizes, both sides, protection modes, and arithmetic-noise prices (e.g. 0.4 + 0.2).

^{Reviewed by Cursor Bugbot for commit 75a0c6d. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 625fc27. Configure here.}