fix(sync): propagate decode errors instead of silently reverting to defaults

[datlechin]

Summary

• SyncRecordMapper.toConnection and toSSHProfile now throws instead of returning optional. Four try? decode sites for sshConfigJson, sslConfigJson, additionalFieldsJson, and jumpHostsJson propagate decode errors through a typed SyncDecodeError instead of falling back to empty defaults.
• SyncCoordinator.applyRemoteConnection / applyRemoteSSHProfile log the failing record name + field, then skip the record. Local state stays intact, so the next push won't overwrite the cloud copy with empty defaults.

Why this matters

Cross-version schema drift caused silent data loss: device on an older app build pulls a record written by a newer build → try? decode of the SSH/SSL config blob fails on an unknown shape → fallback substitutes SSHConfiguration() / SSLConfiguration() → next push from the older device writes empty configs to CloudKit, overwriting the authoritative copy from the newer device.

This is bug B2 from the full-app audit (2 CRITICAL data-loss findings); B1 shipped as #1140.

Test plan

• Manual: write a CloudKit record locally with a malformed sshConfigJson blob, run a pull, confirm the record is skipped (log: Skipping remote connection ...: Sync record decode failed for sshConfigJson: ...) and local connection is unchanged.
• Manual: same for sslConfigJson, additionalFieldsJson, jumpHostsJson (SSH profile).
• Manual: confirm a normal pull with valid records still applies them and advances state.
• swiftlint --strict clean.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.