Tidy up subnormal handling

[joshcarp]

Currently numbers are "normalised" so that there are trailing zeros,
for example 1 is stored as a significand 100000000000000 with an exponent of -15

Generally outputs are normalised from ParseDecimal and arithmetic functions.
Some problems are:

1. Errors may arise when Decimal encodings are parsed in using methods other then strings
2. Arithmetic functions currently have a remove trailing zeros function to avoid integer overflow
   If Decimal64s are stored without trailing zeros arithmetic functions may be faster

[anzdaddy]

In early experiments, I noticed that gcc stores the literal decimal 1 as 1 with an exponent of 0. So either:

1. it doesn't bother to normalise (just store whatever representation is easiest in a given context), or
2. it normalises the other way (use the smallest possible integer that retains all digits).

This may require further experimentation.

The key requirement, however, should be to ensure that the library correctly handles all valid representations of each number, not just one normalised form.