Round in calculation of signal durations to avoid floating point issues

[cbrnr]

Instead of rolling your own tolerant comparison, I would suggest to use math.isclose() instead.

[hofaflo]

Instead of rolling your own tolerant comparison, I would suggest to use math.isclose() instead.

Thanks for the suggestion, I did not know about this! We can use this in EdfSignal._get_expected_new_length.