BigMath.exp overflow/underflow check

[tompng]

BigMath.exp(BigDecimal('1e100000000'), 20) was slow.
Skip calculation when x.exponent >= 21 which always result in overflow or underflow.

BigMath.exp(BigDecimal('0.212e20'), 2) #=> 0.16e+9207043016348938747
BigMath.exp(BigDecimal('0.213e20'), 2) #=> Infinity

BigMath.exp(BigDecimal('-0.212e20'), 2) #=> 0.63e-9207043016348938746
BigMath.exp(BigDecimal('-0.213e20'), 2) #=> 0.0