docs: tighten threat model boundaries

[peg]

Summary

• Replaces the threat model opening with cleaner policy-boundary language.
• Removes roadmap-style phrasing from the legacy patching limitation.
• Removes em dashes from the threat model page.

Tests

• git diff --check
• python3 - <<'PY' ... check for zero em dashes and removed roadmap phrases in docs-site/reference/threat-model.md