FIX: issue #4713 (print special float! not right)

[bitbegin]

No description provided.

[qtxie]

https://en.wikipedia.org/wiki/IEEE_754-1985, All NaNs in IEEE 754-1985 have this format:

sign = either 0 or 1.
biased exponent = all 1 bits.
fraction = anything except all 0 bits (since all 0 bits represents infinity).

A better test for double float NaN and INF:

if d/2 and 7FF00000h = 7FF00000h [ ;-- if biased exponent all 1 bits, it is INF or NaN
    either all [   ;-- if fraction all 0 bits, it's INF
        zero? d/1
        zero? (d/2 and 000FFFFFh)
    ][
        either zero? (d/2 and 80000000h) [prin "1.#INF"][prin "-1.#INF"] ;-- check sign bit
    ][prin "1.#NaN"] ;-- otherwise NaN
]

[bitbegin]

this pr is more simple for me, and can cover all cases

[qtxie]

It's hard for me to understand the code without explaining the rationale of it in a PR, also no comments in the code.

[bitbegin]

https://www.doc.ic.ac.uk/~eedwards/compsys/float/nan.html

[bitbegin]

make this check simply, treat both SNAN and QNAN as NAN. +INF/-INF are special cases, so need process first

[qtxie]

@bitbegin Thanks. In this PR you check those special values by patterns. While I prefer to do the checking by rules. See my first comment. A more lengthy version as below:

1. if all bits in exponent part are 1, it's a special value (NaN or INF)
2. if it's special value, check fraction part
3. if all bits in fraction part are 0, it's a INF, otherwise it's a NaN
4. if it's a INF, check sign bit to distinguish -INF and +INF