Proposal for two new methods on the double class.
The explode method returns a pair of uints [hi,lo], where hi is the high part of the number (sign, exponent, and high bits of the significand) and lo is the low part of the number (low bits of the significand). If the double is a NaN, explode always returns the same representation regardless of the representation of the NaN. This representation is TBD but is platform-independent (QNaN does not have a platform-independent representation).
The implode method takes two arguments hi and lo and returns a double constructed from those parts. If the number constructed is a NaN, the method returns the value of the global NaN constant.
Rationale: picking apart and putting together doubles is useful for serialization, formatting, scanning, etc.
(Something similar is probably useful for decimal.)
