Calculation precision refactor

[stevegeek]

We need to look at ensuring calculations stay within Rational/BigDecimals as float conversions lead to unexpected results. Also I think we should stick to an internal numerical representation and always use it

See:

• Should division of integer measurements return an integer or a rational? #66
• Conversion of ml/h to itself yields floating point/rounding inaccuracy #65
• Cannot convert to decimal #to_d #59
• Conversion from ratio to percentage gives strange results  #61