Update security model opt-out flags to interact with processing models

[joelweinberger]

The security model opt-out flags (i.e. the "unsafe-*" options) have various effects on existing processing models. The subsections should incorporate those effects.