Fixed issue with incorrect temporary target validation

[Erin879]

This fixes the issue described in #4859, by only running the validation code if the temporary target event is actually a target rather than a cancel event.

[Erin879]

Perhaps a better solution would be to remove the code here which turns Temporary Target Cancel events into Temporary Target events, and then that'd avoid the issue entirely. However, I'm unsure on why this code was implemented so I'd rather not include that change in a commit without other people's input.

EDIT: I spent some time reading over the code to try and find any issues with this approach, and the only one I could think of was that previously stored Temporary Target Cancel events would be treated as Temporary Target events. Maybe this could be fixed by removing the code to save cancel events as normal temp targets, and then having two methods (duration of 0, and cancel events) to parse the events, and eventually removing the legacy code? However, this might cause issues with how openaps handles cancelling temp targets, but as I use androidaps I am unsure on that.

[sulkaharo]

The reason we set a new target with 0 duration is simply that the way the data is interpreted is, a new target automatically overrides the previous, and thus a 0 length target cancels the previous target without adding a new target. This is how the data is also interpreted by systems that read Nightscout, so there's no easy way to provide backward compatibility if we add a new event type

[mhaeberli]

When can we expect this to be merged back into master?