docs: refresh memory bank for PropertySpec hardening (os-007)

[TKaltofen]

Refreshes the four memory-bank files that had drifted to a 2025 options/sklearn snapshot so they describe the current PropertySpec model, resolution architecture, supported Python range, and completed PROPERTY_MAPPING hardening work.

Changes

• systemPatterns.md: documents the typed PropertySpec lifecycle (author, parse, bind, match, resolve, materialize defaults, compute), including the raw-dict hard break, framework-enforced default materialization at the compute boundary, and the explicit-None policy.
• activeContext.md: replaces the stale sklearn/options next-steps with the property-mapping-hardening epic status: core hardening complete, os-008 and phase-6 downstream remaining.
• progress.md: updates the working-feature catalog (9 compute frameworks, current feature-group families including the sklearn split and the LLM family, OtelExtender) and the hardening achievements.
• techContext.md: corrects the supported Python range (>=3.10,<3.15), the dependency-free core, and the current tox gate.

Verification

Docs-only change. The tox gate does not exercise memory-bank/*.md (no test reads them; ruff/mypy/bandit are Python-only; mktestdocs targets docs/), so CI is unaffected. Content was verified against the current code and config, and factual-accuracy fixes were applied for the NodeCentrality dependency (numpy, not scikit-learn), the all-optional universal-matcher (a definition-time warning, not a runtime guard), the transitional positional-binding fallback, the PythonDict backend needing no extra, and the parse_name file pointer.

Addresses eng-board os-007 (migrated from mloda#800). Part of the PROPERTY_MAPPING hardening epic.