fix: session filter, short weight cap, futures multiplier in cash checks, CAGR overflow

[quantum-like]

Owning issue

Closes #128

Outcome

5 correctness fixes, one commit each, each with a focused test:

• filter_to_trading_sessions extended the exchange schedule one day before the first bar but stopped on the last UTC date. Overnight sessions such as the CME equity open start on the calendar day before their session date, so a bar in that open was dropped when the frame ended on the prior UTC date. The schedule end is now also extended by one day. Test: tests/test_calendar.py::test_filter_keeps_overnight_open_before_session_date. (3f6eac0)
• TargetWeightExecutor applied max_single_weight with min(), which caps long targets only; a −50% target with a 25% cap stayed at −50%. Short targets are now floored at -max_single_weight. Test: tests/execution/test_rebalancer.py::test_max_single_weight_caps_short_targets. (abc0c4c)
• The next-bar simple cash check valued an order at price × quantity, ignoring the contract multiplier, and then skipped the gatekeeper. One ES contract at 4,000 with multiplier 50 was treated as a $4,000 order, so a $200,000 buy filled against $100,000 of cash. Projected cash now uses price × quantity × multiplier. Test: tests/test_config_wiring.py::test_next_bar_simple_cash_check_includes_contract_multiplier. (e78288d)
• Buying-power reservation for sequential fills had the same gap; the reservation now includes the multiplier. Equity orders are unchanged (multiplier 1). Test: tests/test_config_wiring.py::test_buying_power_reservation_includes_contract_multiplier. (cd49c85)
• cagr() raised OverflowError when a very short window was annualized (e.g. a one-minute move from 100,000 to 101,000), which aborted EquityCurve.to_dict(). That case now returns infinity; finite windows use the same formula as before. Test: tests/test_equity_curve.py::test_short_window_cagr_does_not_overflow. (c9397df)

Compatibility

• No public compatibility impact
• Compatibility impact is documented in the owning issue and migration guidance

Results change only in the cases above (overnight-session bars kept, short targets capped, futures orders sized with the multiplier, overflowing CAGR reported as inf).

Verification

• Tests reproduce the prior failure or verify the requested behavior (each listed test fails on main and passes with its fix)
• Ruff lint and format checks pass (ruff format --check is clean on the changed files and ruff check reports nothing new in them versus main, but I did not run the full pre-commit suite)
• ty passes (not run on this branch)
• Test suite passes: pytest excluding benchmarks: 2198 passed, 9 skipped before; 2194 passed, 9 skipped, 9 failed after.
• Package build passes when applicable (not run)
• Strict MkDocs build passes when documentation is affected (no documentation changes)
• Ecosystem qualification passes (not run)

The 9 failures are the retained-evidence checks in tests/contracts/test_parity_claim_generation.py and tests/contracts/test_real_strategy_runner.py. They hash the engine source files and compare against pins in validation/CORRECTNESS_RESULTS.json and validation/REAL_STRATEGY_RESULTS.json, so any engine source edit changes them. Refreshing those pins needs a re-derived cross-framework run that I could not do, so I left the published evidence files untouched. They are not behavior regressions, but they will need your refresh before merge.

Documentation and release

No documentation changes. The behavior changes above may be worth a line in the next patch-release notes.

Amir