fix: rebase target qty on unfilled position top-ups

[Magic-Jia]

Summary

• Rebase target quantity calculations on the remaining unfilled/open position quantity during top-ups.
• Prevent target-management sizing drift when repeated paper fills update an existing symbol position.
• Preserve paper execution / position lifecycle consistency for S-tier microstructure integration evidence.

Verification

• python scripts/verify.py --changed trading_system/app/portfolio/positions.py --changed trading_system/tests/test_target_management_state.py

  • 1089 passed in 2.86s
  • git diff --check HEAD clean

• python scripts/verify.py --suite full

  • 6733 passed in 185.90s
  • git diff --check HEAD clean

Paper evidence

4h proof bundle:

• /tmp/s-tier-four-hour-performance-proof-retry-I8jwk1/output/summary.json
• /tmp/s-tier-four-hour-performance-proof-retry-I8jwk1/runtime/paper/microstructure-four-hour-performance-proof-retry/execution_log.jsonl
• /tmp/s-tier-four-hour-performance-proof-retry-I8jwk1/runtime/paper/microstructure-four-hour-performance-proof-retry/paper_ledger.jsonl

Observed results:

• 240 rounds
• 360 candidates / allocations
• 304 accepted allocations
• 56 rejected allocations
• 40 paper fills
• 40 unique paper ledger fills
• 6 final open positions
• false-close count: 0
• execution sample health: available
• optimization health: ok
• promotion decision: observe

Promotion status

This is not promoted yet. The current decision is observe, because there are no active recommendations to validate and the latest funnel is risk-off / blocked. This patch closes the top-up target quantity correctness gap and makes the paper evidence loop observable, but longer paper/live-sim evidence is still required before promotion.