perf: reduce repeated work across Binance tabs

[jackhai9]

Long-lived Binance tabs repeatedly sorted native depth without display subscribers, projected unchanged strategy markers, and retained data-panel requests after navigation. Build depth profiles on demand, reuse committed marker presentations, and cancel retired request sessions while preserving background refresh, complete depth validation, signal corrections, and trading checks.

The shared marker change is included in Strategy27, Strategy29 and Strategy31. The affected installers are orderbook 2.7.228, trading-data 1.2.4, CMC-data 0.2.4, Strategy27 0.6.15, Strategy29 0.5.13 and Strategy31 0.1.9. Installation URLs are unchanged.

Browser test synchronization now tolerates Clock transport latency and explicitly separates held responses from the native drawing-discovery window. No production trading deadline changed.

Validation:

• Complete Node suite: 2,616 passed.
• Browser evidence: 381 affected scenarios passed with one worker and zero retries; 239 unchanged passing scenarios reused from the initial run, covering all 620 current scenarios. The initial full run had two diagnosed test-synchronization failures; this is not a single passing full rerun.
• Six builds, generated syntax and source metadata checks, test/UI-copy lint, and whitespace checks passed. Four rendered fixture screenshots inspected; independent read-only reviews passed.
• Controlled counts: unsubscribed depth sorts 402 to 0; unchanged Strategy29 marker projections 4,884 to 444; outstanding historical requests after three rapid symbol activations 21 to 7, then zero on route departure.

The new artifacts have not been installed or loaded in live Chrome. Controlled operation counts do not establish actual CPU or memory reduction. See docs/binance-multitab-performance-review.md for evidence, test history, and remaining live verification.