decode (encode NaN) /= NaN

The IEEE "not-a-number" (NaN) value is not encoded properly, since `encode` uses `decodeFloat`, which is unspecified for NaN (cf. [Prelude](http://hackage.haskell.org/package/base-4.7.0.1/docs/Prelude.html#v:decodeFloat)).

Examples:

``` Haskell
(0 / 0 :: Double) = NaN
(decode (encode (0 / 0 :: Double)) :: Double) = -Infinity
(log (-1) :: Double) = NaN
(decode (encode (log (-1) :: Double)) :: Double) = Infinity
(0 / 0 :: Float) = NaN
(decode (encode (0 / 0 :: Float)) :: Float) = -Infinity
```

Start by reading the Haskell binary package's encode and decode entry points, focusing on the Float and Double path that uses decodeFloat. Reproduce the listed NaN examples, then verify that encoding and decoding NaN preserves NaN for both types.

索引モデルが issue の本文から書いたものです。