[codex] Support TCPViewer session export and import

[NghiaTranUIT]

Summary

Adds .tcpviewsession as a ZIP-backed TCPViewer session format with readable JSON sidecars, an internal capture.pcapng, de-duplicated client records, annotations, document-local state, and packaged client icons.

The import path now handles malformed per-flow records gracefully: unsafe packages still fail at the package boundary, but malformed packet rows are skipped, valid flows continue importing, and TCPViewer shows an alert with imported/skipped flow counts only when skipped flows are present.

What Changed

• Added session format models plus export/import services for .tcpviewsession packages.
• Added File menu export flow and document type registration/icon assets.
• Preserved document-local pins, saved packets, custom filters, source-list state, table state, comments/colors, client metadata, and app icons without mutating global user state.
• Added client de-duplication and document-local icon resolution for packet table/source list rendering.
• Hardened session import against malformed ZIP paths, unsupported schema, duplicate IDs, dangling refs, bad JSONL rows, invalid ordinals, packet ID mismatches, and missing optional icons.
• Mapped session packets to raw capture.pcapng packets by captureOrdinal, so skipped rows do not shift inspection/export onto the wrong packet.

Validation

• git diff --check
• xcodebuild -project TCPViewer.xcodeproj -scheme TCPViewer build CODE_SIGNING_ALLOWED=NO TCPVIEWER_DEVELOPMENT_TEAM=ABCDE12345 TCPVIEWER_BUILD_KEY=local-build
• xcodebuild test -project TCPViewer.xcodeproj -scheme TCPViewer -destination platform=macOS CODE_SIGNING_ALLOWED=NO TCPVIEWER_DEVELOPMENT_TEAM=ABCDE12345 TCPVIEWER_BUILD_KEY=local-build

The exact default build/test commands still fail locally because this machine does not have a valid Apple Development identity for the SMJobBless requirement generation step: Invalid Apple Development signing identity for SMJobBless requirements: Sign to Run Locally.

Disclosure

This PR was substantially assisted by AI and reviewed through local build/test runs before publishing.