Use warning message in quick close

[t-bast]

Once #834 have been accepted to the spec, we should use a warning in quick close instead of failing the connection.

[Roasbeef]

If a warning message is added (again without a feature bit??), and you send this to a peer that maybe understands the new co-op close TLV field (since there's actually no feature bit, so you just need to guess), then how do you know they actually even support this feature?

This is the whole reason we added dependencies to feature bits: we can make this relationship explicit, namely that the new co-op close TLV depends on understanding of the warning message (as it's used to signal an error case).

[t-bast]

you send this to a peer that maybe understands the new co-op close TLV field

That condition is nested below - if the message contains a fee_range: so your peer does understand this TLV field.

Again, your argument is true in general but in this specific case it's completely unnecessary.

[TheBlueMatt]

Additional context here may be useful: the warning message is only used to indicate "hey, we couldn't come to agreement!", but that's also implied by the fact that, well, you didn't see a useful response. No matter if the peer responds or not, you have to eventually time-out the negotiation and force-close (or let the user do that, which they'll presumably do if we take forever to negotiate fees), which is the same as today. The warning message is just a helpful bit of context for users who read their debug logs, but the response to the condition that is warned on should really be the same either way.

[vincenzopalazzo]

ack db636f0

[rustyrussell]

Ack, easy!