Enrich Tried all MXs defer with per-MX failure reasons

[gamalan]

Context

When delivery exhausts all MXs, Haraka currently defers with a generic:

Tried all MXs example.com

This makes triage hard — logs have per-MX errors but the queue reason /DSN / bounce doesn't. Operators have to correlate Failed to get socket /Ongoing connection failed log lines. When mxlist is empty (e.g resolve_mx_hosts returns [] or DNS edge case), the reason is even more opaque.

Proposal

Include collected per-MX failure details in the final Tried all MXs<domain> defer, and propagate that into the DSN via extend_rcpt_with_dsn.

Before:

Tried all MXs example.com

After (with failures):

Tried all MXs example.com: mx1.example.com:25 connect ECONNREFUSED;
mx2.example.com:25 socket timeout waiting on connect

currently i implement on our internal system https://github.com/gamalan/Haraka/tree/feat/enrich-tried-all-mxs.

Is this would be helpful for other? Can I make PR?

[msimerson]

I can see that being useful to folks that don't feed their logs into an indexer. Sure, make the PR.