Add Network migration readiness audit

[rchatham]

Summary

• Add the authoritative readiness audit for graduating .networkFramework from experimental opt-in to stable opt-in, then default, and eventually removing MultipeerConnectivity.
• Inventory completed capabilities and accepted non-parity against the current implementation and tests.
• Classify optional PR Add Network migration readiness audit #50+ work by when it becomes required: before stable, before default, only before removal, product-dependent, or optional.
• Record release/versioning constraints and a risk register.
• Reconcile and cross-link the existing migration plan, PR plan, backend guide, and README.

Decision

The current Network backend is suitable for bounded experimental opt-in evaluation. It is not ready for stable/default status or MultipeerConnectivity removal. No runtime merge blocker was found for the current stack, so this PR is documentation-only and does not change the default backend.

Key gates

• Before stable: individual peer authentication, public Network failure/permission observability, discovery/idle resource bounds, soak/churn testing, and recorded physical-device/topology validation.
• Before default: at least one stable opt-in release cycle, a secure replacement for the current unauthenticated default initialization path, complete migration guidance, and rollback criteria.
• Before MC removal: deprecate/replace MC-specific public surfaces and either implement or intentionally remove stream/resource behavior in a breaking release.

Verification

• swift test — 113 tests passed.
• xcodebuild test -project PeerConnectivity.xcodeproj -scheme PeerConnectivity -destination 'platform=iOS Simulator,name=iPhone 16,OS=18.3.1' -configuration Debug — 113 tests passed.
• git diff --check and local Markdown link validation — passed.
• Correctness review: no merge-blocking findings.
• Security review: no merge-blocking security findings.

Risks

The current group-PSK model does not authenticate individual peers; Network startup/permission failures are not consistently surfaced; discovery is unbounded and connections have no idle timeout; and simulator/loopback tests do not replace physical-device validation. These are explicitly recorded as pre-stable gates. No security behavior changes in this documentation-only PR.