inf and nan floats doesn't encode_value

Imported from http://bugzilla.roxen.com/bugzilla/show_bug.cgi?id=1919

Reported by Mirar , Idonex Heavy Industries mirar@roxen.com

> Math.nan; decode_value(encode_value(Math.nan));
Result: nan
Result: -2.000000
> Math.inf; decode_value(encode_value(Math.inf));
Result: inf
Result: -2.000000

Suggested solution: Encode floats using IEEE standard (suitable size, 64 bits if pike is using doubles). Let people who need backwards-compatible encode_value use a backwards_compatible_encode_value function or similar.