Make PacketProcessor safe without prior SetMsgHandler

[jclark]

Follow-on cleanup from issue #210

Initialise mh to DefaultHandler{} internally in each PacketProcessor so it is never nil. Remove mh != nil guards in dispatch paths (ubx Dispatch lines 156/212, ubx flushNavEpoch line 87, casic, as). Update PacketProcessor interface comments to reflect the new contract: SetMsgHandler is optional; processors are safe to use without it. Add tests to every packet processor that they can process packets with no prior SetMsgHandler call. Once verified, remove the DefaultHandler install from satpulsetool (the internal default suffices).

[jclark]

Once the `mh` field in each `PacketProcessor` is never nil, `FlushNavEpoch` will always return a non-nil `MsgHandler`. After that lands, the nil-handler scan in `NavEpochManager.flush` can be simplified to just use `items[0].mh` directly.