Function 'round_err' works now correctly for cases such as

- e = round_err(0.1078); -> e.value = 0.11 (and not 0.1)
- e = round_err(0.6); -> e.value = 0.6 (and not 0.6000000000000001)
- e = round_err(0.4); -> e.value = 0.4 (and not 0.5)