Until this commit, `int` and `float` were special types in Pint
(just as they are in Python). Numbers were parsed from strings
as int/float from the definition files and from user provided
strings; and exponents of units were also stored as int/float.
This commit change this by adding a new argument (`non_int_type`)
to classes and methods. It indicates how numeric values will be parsed
and defaulted. Any numerical class can be used such as `float` (default),
Decimal, Fraction.
This argument will be found in the following places
1. UnitRegistry: used for parsing the definition files and any
value provided as a string.
2. UnitsContainer: used to compare equality with strings, multiply and
divide by strings (which is equivalent to parse the string)
3. All methods OUTSIDE the UnitRegistry/Quantity that can parse
strings have a `non_int_type` argument. (e.g. Definition.from_string)
Tests have been added for by duplicating most cases in `test_quantity.py`.
(Some tests have been deleted such as those dealing with numpy.)
The new file `test_non_int.py` run the tests for `Decimal`, `Fraction` and
`float` (which is redundant but is kept as a crosscheck for the implementation
of this testsuite)
BREAKING CHANGE: `use_decimal` is deprecated.
Use `non_int_type` keyword argument when instantiating the registry.
>>> from decimal import Decimal
>>> ureg = UnitRegistry(non_int_type=Decimal)