FAILEDREADFREQUENCY is overwritten

[lucasnz]

In updateStatus(), if a read error occurs the function sets:
_nextUpdateDue = millis() + FAILEDREADFREQUENCY;

Reads (almost??) always occur after a sendCommand(). This function sets:
_resultRegistersDirty = true;

This means that _nextUpdateDue gets overwritten, in loop():
if (_resultRegistersDirty) {
_nextUpdateDue = millis() + 200; // if we need to read the registers, pause a bit to see if there are more commands coming.
_resultRegistersDirty = false;
}

This can result in repeated update requests being sent to the spa in the following scenarios:

1. TX or RX wires aren't connected properly.
2. There is an error in the code parsing the spa response.

Skip's Discord comment on the issue:
So the 200ms is a general delay, the idea being is that if there are a number of writes going to the spa then we don't want/need to be doing a read after each, you can send the writes and then do a single read.

In the original implementation I set up a fifo queue and only did a read when the queue was empty, I did away with this as it added complexity that I didn’t think that we needed

Failed read frequency is meant to be the time between failed RF commands, if we don't get a good result when parsing the RF array then we try again in 1 sec.

In hindsight my design is flawed as it assumes linear processing but there are non-linear elements to the code that may introduce “wriggles” such as the callback from a MQTT subscription.

Possible a duplicate of #51

Do we need to rethink (and reintroduce) the FIFO queue idea? Do we need to add code to limit the number of times we retry with a 200ms delay and then start backing off?