fix: use round_down for market order price

[skyc1e]

get_market_order_amounts uses round_normal for price, but the TypeScript client uses roundDown in the equivalent getMarketOrderRawAmounts (helpers.ts:200).

round_normal can round up (e.g. 0.555 → 0.56 with 2 decimal precision), producing a worse effective price for BUY market orders — fewer shares per dollar.

This changes line 88 of builder.py from round_normal to round_down for market orders only. Limit orders (get_order_amounts) are unchanged and correctly use round_normal in both clients.

Includes a regression test using a price that diverges between the two rounding methods.

Closes #323

---

Note

Medium Risk
Changes market order price rounding, which directly affects computed maker/taker amounts and effective execution price. Risk is contained to market orders but could impact downstream order validation and user outcomes if rounding assumptions differ.

Overview
Market order construction now rounds the input price down (via round_down) in get_market_order_amounts instead of using round_normal, preventing cases where mid-tick prices round up and produce a worse effective price (notably for BUY market orders).

Adds a regression test (test_market_order_uses_round_down_for_price) that exercises a price (0.555 with tick_size="0.01") where round_normal and round_down diverge, ensuring the builder matches the expected (TypeScript-aligned) rounding behavior.

^{Reviewed by Cursor Bugbot for commit ac9827a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

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

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

^{Reviewed by Cursor Bugbot for commit ac9827a. Configure here.}

[cursor]

SELL regression test missing price rounding assertions

Low Severity

The SELL half of test_market_order_uses_round_down_for_price only asserts that side == UtilsSell, without verifying that round_down was actually used for the price — unlike the BUY half, which explicitly checks effective_price against both round_down and round_normal results. This means a future regression that changes SELL price rounding back to round_normal would not be caught by this test, defeating the stated purpose of a regression test for issue #323.

 

^{Reviewed by Cursor Bugbot for commit ac9827a. Configure here.}