Strange behavior with LS240 channels disabled

[BrianJKoopman]

@ttsan2521 found some strange behavior while trying to disable channels in one of the 240-8p's on site.

1. Disabled channels readout all the time.

Channel 4 on LSA231C was reading out zero, but then had strange readings on it overnight, it wasn't disabled. It still reads zero after disabling, so it's difficult to distinguish between "disabled" and "reading 0".

While investigating this it seems like after disabling a channel, a .get_reading(unit='K') still returns the last known temperature value, while .get_reading(unit='S') returns 0. This is strange. It would be better if disabled channels just didn't publish any data.

2. Disabling a working channel then restarting the agent shows abnormal readings. BE-OT is the channel in question here, reading out ~4.6K. In this plot it was reading out normally, the Agent was stopped, the BE-OT channel disabled, then the agent restarted. Upon restart BE-OT reads ~22K, despite being disabled. Then the agent is stopped, the channel re-enabled and the agent restarted. It resumes reading out ~4.6K. I'm not really sure what's going on in this case, since directly querying the device yields the stale ~4.6K readings as described in the behavior of 1.

I think the path forward is splitting up the data being published so that we don't get a changing block structure, then tracking which channels are enabled/disabled and not publishing values for them. It'd also be good to figure out behavior 2 here, but if we don't return anything for disabled channels, it's maybe less of an issue.