Refactor AddExponent overflow/underflow check

[tompng]

Just a refactor for BASE_FIG=9 (DECDIG=int32_t) and SIGNED_VALUE is 64bit

Fix this bug

RUBY_PLATFORM
# => "arm-linux-eabihf" (arm32v7/ruby docker image)
x=BigDecimal("0.1e-214748364")
y = x*x*x*x*x*x*x*x*x*x*0.1
# => 0.1e2147483646 # should underflow

Fix this bug for BASE_FIG=4 (DECDIG=int16_t)

x=BigDecimal("0.1E-922337203685477579");
y = x*x*x*x*x*x*x*x*x*x
#=> 0.1e-9223372036854775799
y.precs
# => [4, 12] # by undefining HAVE_INT64_T
z = y*0.00000000001
# => 0.1e9223372036854775806 # exponent underflow. result should be 0.0

# Reason of the bug
# z is 0.0001_0000_0000_0000e-9223372036854775808
# real->exponent is 2305843009213693952 and real->exponent * BASE_FIG does not underflow
# but the actual exponent returned to ruby is -9223372036854775811. This is smaller than SIGNED_VALUE_MIN.

This bug does not happen when BASE_FIG=9 and SIGNED_VALUE is 64bit because SIGNED_VALUE_MIN + BASE_FIG - 1 == -9223372036854775800 is a multiple of BASE_FIG == 9