fix: use broker close tick for pinned POOC short exits

[luisleo526]

Evaluate a reissued full short exit against the broker's rounded close under ordinary process_orders_on_close. The previous raw-close comparison could fill a limit one bar early or a stop one bar late. Preserve the original full-percent request and ignore only exit siblings whose parent never filled in the current position; active competing orders retain the existing behavior.

Validation on engine 113ec87258f02cfee12db57f89b5631b0bb4bda9 with codegen 3fd97fe28abd7b191cb376d9f5db4e056fcfae4b:

• Cloud Run measured all 4,190 probes in 72/72 successful cases. Hariss on Ford moves from strong to excellent: count difference 1 to 0, 100% match, and zero canonical P90 errors. Its 742 entry/exit rows match TradingView on side, time, price and quantity. All other 4,189 probes retain their canonical metrics; zero regressions, missing measurements or new moderate results.
• Gate pineforge-pr-gate-2vzwg: PASS against baseline 86049406d8f675d7ad237f519d1a023ae21f6096, target score +1, zero regressions or tolerated exceptions across 704 hard probes. Result: 4,157 excellent / 32 strong / 1 moderate.
• Fresh all-target build, 210/210 unit tests, 183 focused checks, 27 unchanged parent exclusion panels, C ABI and freshness checks passed. Independent Grok 4.6 review of this exact candidate is GREEN, P0/P1/P2 = 0.

The change is limited to the pinned older single-short-position exit shape. Order levels, fill calculations, grading metrics, thresholds, tolerances, verifier, harness, tapes and population are unchanged.

Candidate snapshot: 0099071a96ee899f19f2ac5c655b27d7161c51348d359866076dbaf2d5bb7c5d (788,857 bytes).