MQTT SUB payload stored as ByteArray object reference instead of decoded text

What happened

ServerMqttManager received MQTT messages and passed the payload to the message handler using message.payload.toString(). In Kotlin/JVM, ByteArray.toString() returns the JVM object reference string ([B@1cb85350), not the UTF-8 text content. Every inbound MQTT SUB message therefore stored a garbage heap-address string into the snapshot and connected DataPoints, making the full external publisher → MQTT SUB → DataPoint → automation path produce junk downstream. The bug was visible in logs: the line above the broken call already decoded correctly with String(message.payload) for logging purposes, but the handler invocation did not.

Introducing commit: ad3d0b6c7 mqtt working.

Fix

• server/src/jvmMain/kotlin/.../mqtt/ServerMqttManager.kt line 66: changed message.payload.toString() → String(message.payload, Charsets.UTF_8).

Prevention

• ByteArray.toString() in Kotlin/JVM never produces the byte content — it always produces the JVM object reference. Prefer String(byteArray, Charsets.UTF_8) (or .decodeToString()) wherever a ByteArray must be interpreted as text.
• When a log line already decodes correctly (String(message.payload)) but the adjacent production call does not, that is a strong signal the log was written first and the handler call was copy-pasted without the same care.
• Add a regression test that constructs a MqttMessage with known text, round-trips through the decode path, and asserts equality — no broker connection required.