Handle Network backend reconnects

[rchatham]

Summary

Adds a bounded Network backend reconnect E2E scenario and fixes the reconnect issues it exposed. Also stabilizes the existing multi-peer broadcast E2E test by retrying the broadcast after connections settle.

Changes

• Add testNetworkBackendReconnectsAfterPeerRestarts.
• Verify Alice observes Bob connect, disconnect after Bob stops, then reconnect when a replacement Bob starts with the same service type/security.
• Treat remote receive completion/errors as connection cancellation so the coordinator emits disconnect cleanup for the remaining peer.
• Keep the existing automatic refresh behavior MultipeerConnectivity-only; Network transports keep their listener/browser running after disconnect so they can discover and reconnect without trying to restart cancelled NWListener/NWBrowser instances.
• Stabilize the three-peer broadcast E2E by sending a short retry burst after Alice has connected to both peers and allowing duplicate receives to avoid over-fulfillment failures.

Verification

• swift test --filter NetworkPeerLoopbackTests/testNetworkBackendReconnectsAfterPeerRestarts — passed
• swift test --filter NetworkPeerLoopbackTests — 6 tests passed
• swift test — 108 tests passed
• xcodebuild test -project PeerConnectivity.xcodeproj -scheme PeerConnectivity -destination 'platform=iOS Simulator,name=iPhone 16,OS=18.3.1' -configuration Debug — 108 tests passed
• GitHub Actions pull_request CI — Swift Package Tests and Xcode Project Tests passed
• GitHub Actions push CI — Swift Package Tests and Xcode Project Tests passed
• Correctness reviewer — no merge-blocking findings
• Security reviewer — no merge-blocking security findings

Notes / limitations

• This PR focuses on one peer-restart reconnect path. More aggressive churn/stress coverage should stay in separate slices to avoid broadening real Bonjour/Network.framework test flakiness.

Stack context

Stacked on PR #36 (feature/network-migration-multipeer-e2e). This is the reconnect E2E coverage and fix slice of the Network migration stack.

[Copilot]

Pull request overview

This PR adds end-to-end coverage for Network backend disconnect/reconnect behavior (peer restart) and adjusts Network transport/session handling to properly emit disconnect cleanup, while also stabilizing the existing multi-peer broadcast E2E test to reduce flakiness.

Changes:

• Add a bounded Network backend reconnect E2E test (testNetworkBackendReconnectsAfterPeerRestarts).
• Update Network receive handling to treat remote receive completion/errors as disconnect signals for cleanup.
• Stabilize the existing multi-peer broadcast E2E test by retrying the broadcast and tolerating duplicate receives.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
Sources/PeerConnectionManager.swift	Limits automatic refresh-on-disconnect behavior to MultipeerConnectivity backend to support Network backend reconnect behavior.
Sources/NetworkPeerTransport.swift	Adjusts Network receive loop behavior to emit disconnect/cancellation cleanup when the remote side completes/errors.
PeerConnectivityTests/NetworkPeerLoopbackTests.swift	Adds reconnect E2E scenario and stabilizes multi-peer broadcast E2E via retry + duplicate-tolerant expectations.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[rchatham]

Superseded by consolidated PR #42. No commits were discarded; this branch is preserved as a recovery and reference point. Addressed review findings were replied to and resolved before consolidation.