Import and header adjustments

[sstendahl]

Makes the following adjustments to importing and header handling:

• Stops looking for headers when data is detected. This makes it such that comments and other data at the bottom are just ignored
• Only add X and Y data if both data can be added, otherwise skip that row. This prevents errors when one of the data points was valid but the other one was not
• Use the getters and setters for the labels, this was broken when we moved the item class to Vala, the relevant code was still trying to assign the attribute directly.
• Do not add X and Y data if either of them returns None. This still parses as a float, but leads to errors when doing fitting or limit optimisations
• Just use the delimiter to detect headers, stop detecting whitespaces as first choice by default. As I feel using two different heuretics to detect headers will lead to more issues than it solves, and you'd pretty much always use the same delimiter anyway
• Remove headers from the axis if the associated item is removed. This is because otherwise adding a dataset with headers, and then deleting it, and then adding another dataset, will put that new dataset in the top/right axis, even though it's the only dataset there. That is not intuitive behaviour and not what anyone would want probably.

Solves issues #674, #675, #677