Adjust modulus/remainder calculation for infinity arguments

```ruby
require 'bigdecimal'
bd = BigDecimal("4.2")
bd.remainder(BigDecimal("+Infinity")) # => NaN, should be bd
bd.remainder(BigDecimal("-Infinity")) # => NaN, should be bd
```

See https://bugs.ruby-lang.org/issues/6120

I've submitted a pull request to ruby to fix the issue with Integer/Float, but bigdecimal should be handled separately.