Fix float precision

[rwols]

Replaced

std::numeric_limits<T>::digits10 + 1

with

std::numeric_limits<T>::max_digits10

everywhere.

This fixes #469

Understandably some tests will start failing now

[rwols]

6ebfa37 should fix the failing unit tests. I tried using

Emitter emitter;
emitter.SetFloatPrecision(3);
emitter.SetDoublePrecision(3);

but this only works for the very first float (or double) that gets emitted -- the floats that follow use the default precision. So I had to replace things like

1.01 --> 1.009999999998