Remove invalid debug_asserts in peer_connected/peer_disconnected

[tnull]

In commit 7d56270, we decided to call peer_connected/peer_disconnected on the previously-added send_only_message_handler.

However, in practice send_only_message_handler will be implemented via ChannelManager, having us call ChannelManager::peer_connected/peer_disconnected at least twice (also for ChannelMessageHandler), resulting in hitting the A peer shouldn't be connected twice and Unconnected peer disconnected debug assertions.

Here, we simply drop these outdated assertions.

[ldk-reviews-bot]

👋 I see @TheBlueMatt was un-assigned.
If you'd like another reviewer assignment, please click here.

[tnull]

(Alternatively the commit mentioned in the OP could be reverted, but this approach here seemed cleaner to me)

[TheBlueMatt]

My understanding is that the send_only_message_handler will be the ChainMonitor, not the ChannelManager.

[tnull]

My understanding is that the send_only_message_handler will be the ChainMonitor, not the ChannelManager.

Ah, good point indeed, seems to stem from a mixup in the LDK Node upgrade PR. However, the fact that I was able to plug-in ChannelManger without issue seems to reveal a weakness in the chosen API/type design? Maybe send_only_message_handler shouldn't be a generic BaseMessageHandler then afterall, since we expect it to be a specific type. I think we might have discussed that on #3623.

[TheBlueMatt]

Hmm, yea, interesting question. Not exactly sure what the right API would look like, I suppose there's not much harm in defining a new trait that is an empty supertrait of BaseMessageHandler.

[tnull]

Hmm, yea, interesting question. Not exactly sure what the right API would look like, I suppose there's not much harm in defining a new trait that is an empty supertrait of BaseMessageHandler.

ACK see #3922

Closing this.