Implement BigMath.frexp and ldexp with exponent of 10

[tompng]

BigMath.frexp(BigDecimal(123.456))
#=> [0.123456e0, 3]

BigMath.ldexp(BigDecimal("0.123456e0"), 3)
#=> 0.123456e3

These methods doesn't take a precision. It just construct or deconstruct BigDecimal from/to fraction part and exponent part without precision loss.

frexp stands for fraction + exponent, ldexp stands for load exponent. It's not limited to binary number.
Math.frexp and Math.ldexp calculates with exponent of 2
BigDecimal version of frexp and ldexp will calculate with exponent of 10.

These two method could be considered an public API version of internal-use BigDecimal#_decimal_shift