Partial exits: exits share each trade's quantity in placement order

[alexgrover]

Builds on #431 (merged): it uses its path helpers and per-trade trailing state.

Several strategy.exit orders on one entry, measured on TradingView (BINANCE:BTCUSDT 1h, CME ES, Pine v6) with 11 new partial-exit probes.

Changes

Exits share each trade's quantity in placement order (exitCapacity). Each exit used to close up to its own qty / qty_percent of the position, so a full-size stop placed after a 50% target closed everything. On TradingView, each exit takes its share out of what the exits placed before it, and still working, have not taken. The share is its qty_percent of the trade's entry quantity, its absolute qty (oldest trades first), or all the rest.

• 50% target first, then a stop: the stop closes half (pe01);
• the stop placed first takes everything, and the target never fills (pe02);
• 30% / 30% / rest (pe03); 60% + 60% gives the second 40% (pe04); qty = 3 against 1 open takes all (pe05);
• an exit that re-calls an id keeps its place, also after it filled on one of the trades (pe06, pe11);
• on whole contracts, 50% of 1 contract takes the contract (x10 ES 1D);
• under pyramiding, the reservation is per trade: an earlier 50% exit holds half of each trade (pe11).

Exit orders are taken in path order. When two exit orders trigger on one bar, the one whose first leg the path reaches first fills first (pe01/03/04).

strategy.close with qty / qty_percent keeps the exits working. Only a full close cancels the entry's pending exits. After a 50% close, the 50% target still fills on that bar (pe07).

A reversal takes its exits with it. Exits without from_entry that were placed while a position was open die when a reversal closes that position, as they already did when it went flat. The new position gets the exits placed for it (pe11).

Tests

tests/namespaces/strategy/partial-exits.test.ts (8 tests): the placement-order cases, over-allocation, qty over the position, whole contracts, a re-called exit after a fill on an older trade, a partial strategy.close, and the reversal. 7 fail on the base branch; "stop placed first takes all" guards existing behavior.

Results against TradingView (compared with dev, with #431 included)

• Partial-exit probes: pe01–pe10 identical (all on dev diverged from trade 0–3); pe11 first divergence trade 0 → 43.
• Controlled cases 10 and 10 @d, x10 BTC 1h and x10 ES 1D: identical. x10 AAPL 1D / 1h: trade 0 / 3 → 2 / 9 (AAPL off-tick prices).
• Public strategies: quantumforex-sniperfusion becomes identical.
• Premium backtesters: as Trailing stops and fills along the bar path #431.
• No run gets worse.
• Unit suite: all committed tests pass. Oracle suite: 0 regressions.

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: daa806eb-38c9-4774-83ff-0ecc77197fbc)