For example: casting an xsd:Integer to string requires the following in most cases
- un-inline the integer value into a int64_t (this is true for most integers)
- create a
boost::multiprecision integer out of it
- call the to string function for the
boost::multiprecision integer
In theory it would be more optimal if we did not construct the boost integer, instead just converting the uninlined int64_t to string.
Not sure on the implementation yet.