FlyingSocks: resolving device tree hangs with large device trees and TCP

[lhoward]

With TCP and a large recursive resolution of the device tree, we see an error:

% swift run ocacli -h 127.0.0.1 -f enableTracing -r
Building for debugging...
[1/1] Write swift-version--58304C5D6DBC2206.txt
Build of product 'ocacli' complete! (0.22s)
2024-11-28T15:29:58+1100 trace com.padl.SwiftOCA : [SwiftOCA] _updateConnectionState: notConnected => connecting
2024-11-28T15:29:58+1100 info com.padl.SwiftOCA : [SwiftOCA] connected to oca/tcp/127.0.0.1:65000
2024-11-28T15:29:58+1100 trace com.padl.SwiftOCA : [SwiftOCA] _updateConnectionState: connecting => connected
2024-11-28T15:29:58+1100 debug com.padl.SwiftOCA : [SwiftOCA] refreshing device tree
2024-11-28T15:29:58+1100 trace com.padl.SwiftOCA : [SwiftOCA] monitor error: pduTooShort
2024-11-28T15:29:58+1100 trace com.padl.SwiftOCA : [SwiftOCA] _updateConnectionState: connected => connectionFailed
^C

Note we don't see this with UDP as ``resolveActionObjectsRecursive()` does client-side recursion for datagram connections, thus limiting the packet size.

I had thought it might be a MTU issue (as loopback can use larger frames) but, even if it is, it doesn't appear to occur on Linux.