Fix determineUnitForFormatting floating point error

[benmccann]

No description provided.

[nagix]

I was also looking into this problem today.

[simonbrunel]

@benmccann can you please detail the issue (why we need +1 and the 0.1 epsilon) and ideally a codepen that reproduces it?

[nagix]

The reason of -1 is the number of ticks includes the ticks at start and end, but we want to divide (max - min) by the number of tick intervals.

[nagix]

The floating point error occurs when the time difference in milliseconds is converted to months or years because the length of a month or year varies.

[benmccann]

@simonbrunel I'm assuming you don't need me to create a codepen since you removed the "needs test case" label. The test I wrote reproduces it though and fails without this change

[simonbrunel]

Correct, I was a bit confused about the - 1.1 but @nagix explanations helped.