Switch all Small Basic numerics to a decimal64-based 'Number' type

[McSherry]

This issue stems from a Yammer discussion.

Small Basic should switch to using the IEEE decimal64 format to represent its numbers. This would:

• Retain the intuitive ability of .NET decimal to represent decimal fractions (as opposed to float and double's binary fractions)
• Avoid any potentially unintuitive (to beginners) side effects of the various type conversions Small Basic does
• Provide opportunities to increase maths performance over .NET's non-standard decimal type, which has been a source of complaints in the past
• Expand the range at the cost of precision, from 10^±28 with 34-digit precision (.NET decimal) to between 10⁻³⁸³ and 10⁺³⁸⁴ with 16-digit precision. Most learners would probably value added range over being able to represent the weight of an electron in grams (which, FWIW, is 9.11×10⁻²⁸).
• Improve Small Basic's portability to constrained platforms—a platform such as a BBC micro:bit might use decimal32 instead, which retains all the same semantics with more limited range and precision. If the decimal64 implementation were limited to the range of decimal32, the switch would only incur a reduction in precision, easing understanding.