fix: Structured API Layers PR review (suggested fixes)

[fcecin]

Description

This PR implements some code suggestions for the Structured API Layers PR that I came up with while reviewing the PR.

Merging this PR is not needed. It just one way to address the points in my code review (see below).

Can also cherry-pick these, or do something different/custom in the code, or leave unaddressed and just close this.

PR Review:

• MessagingClient.new kills the host process. It quit(QuitFailure)s on a recoverable SendService.new failure (a node with neither relay nor lightpush). Via the FFI start path (node_api.nim -> mountMessagingClient) that aborts the embedding app instead of returning an error, and it left the initMessagingClient(...).valueOr branch dead.
• LogosDelivery.stop() abandons teardown on the first failure. An exception in one component early-returns, so e.g. a throwing channel-manager teardown leaves the messaging client and the libp2p node running (leak).
• getNodeInfo() nil-derefs. Unlike kernel()/messaging()/channels() it doesn't guard self.waku.isNil() -> crash if called before start / after stop.

Changes

• MessagingClient.new: return a Result via a new newMessagingClient factory instead of quit(QuitFailure) on a recoverable SendService failure; no longer aborts the FFI host (restores master's behaviour). Callers (facade, FFI mount, tests) updated to the factory.
• LogosDelivery.stop(): best-effort run-to-completion; stop and nil every component even if a prior one fails, aggregating errors (was: early return that leaked the messaging client and the node).
• LogosDelivery.getNodeInfo(): nil-guard self.waku like the sibling getters (was: nil-deref before start / after stop).

[github-actions]

You can find the image built from this PR at

quay.io/wakuorg/nwaku-pr:3966

Built from fe7122c

[Ivansete-status]

@fcecin - the changes looked correct to me, conceptually. However, given that we are not merging POC in the short term, we may need to submit a separate PR based on master, instead.

[fcecin]

@fcecin - the changes looked correct to me, conceptually. However, given that we are not merging POC in the short term, we may need to submit a separate PR to based it on master, instead.

Yep this PR merges into Zoltan's PR, it's just a "review PR." He can drive it however he likes as part of the rewview, close etc. If he ever merges his PR this is closed regardless.

[fcecin]

Actually, let me mark this as draft so it doesn't cause confusion.