Need to be consistent on not Dos-y behaviors from random nodes

[ariard]

See further #151

[TheBlueMatt]

Specifically, the comment at #151 (comment).

[SWvheerden]

What is the consistent behaviour that isa desired here for DOS like behaviour from nodes?
I saw in #151 that it talks about the error action none, but MsgHandleErrInternal::send_err_msg_no_close is filling in the action with a some<>

[TheBlueMatt]

After that comment, in working further on things, I think a reasonable approach would be to replace HandleError in Channel returns (and some places in ChannelManager) with some Channel-specific return value which just says "SendErrorMessage" and then have ChannelManager handle_error! map it to DisconnectPeer or SendErrorMessage based on whether this will close the last of our channels with that peer or if we have some other channel with them just SendErrorMessage.

[TheBlueMatt]

This would give us an opportunity to clean up the stupidly long error-return lines by deduplicating the err string and not having so many optitons to pick from.