Fix edgecase segfault of BigDecimal#remainder

[tompng]

There is a possibility where coerced remainder returns Qnil.
NIL_P(BigDecimal_divremain(self, r, &d, &rv)) doesn't mean d and rv has a calculated value.

This pull request is kind of a refactoring, but also fixes this segmentation fault.

obj = Object.new
def obj.coerce(*)=[self, self]
def obj.remainder(*) = nil
BigDecimal(1).remainder(obj) #=>  [BUG] Segmentation fault at 0x0000000000000020

[tompng]

Same interface as BigDecimal_DoDivmod

        if (BigDecimal_DoDivmod(self, b, &div, &mod)) {
            return BigDecimal_to_i(VpCheckGetValue(div));
        }
        return DoSomeOne(self, b, rb_intern("div"));

[tompng]

I don't think using Qfalse and Qtrue instead of false and true, but this pattern is frequently used in bigdecimal.c even if the function name starts with BigDecimal_.
But this pattern is used in is_zero, is_one and BigDecimal_DoDivmod.
I think it's better to refactor in a separate pull request