implements cast methods for (u)int64 and (u)int32

[dwlnetnl]

See discussion #134.

---

This change is

[philhofer]

You've added some additional same-type casts here, but I think #134 was more about cross-type casts that succeed because the 'value' of the number is the same (e.g., the bits 0x0000000000000000 and 0x8000000000000000 are both the value 0 under IEEE-754, so casting those to any integral type would be legal. Similarly, integral numbers that fit in a floating-point mantissa could be legally casted to floats as well.)