When dividing BigDecimal by a Float, the Float is converted to
Rational “for higher precision,” and then divides the numerator
by the denominator.  However, when doing that division, `div19`
was being used which does a `floor` on the returned float.  This
commit replaces the call to `div19` with `op_div`.
This commit also addresses issues with return type of the result,
which changes from version to version in Ruby.
See https://github.com/rubyspec/rubyspec/pull/220 for tests.

Use java.math.BigDecimal to compute rational and float value.

Make test applicable for 1.9/2.0 only.

Code cleanup.

A bit of housekeeping, adding whitespaces after commas, and around
equals.

Return Float when BigDecimal is divided by Float in 1.9.

Recent versions of 1.9.3 now exhibit the same behaviour as 2.0.