Every operations with NaN should return NaN

[maxonfjvipon]

Accordingly to IEEE 754 NaN should be a result of next operations:

1. Every math operations with NaN as argument
2. 0 / 0
3. ∞ / ∞
4. 0 * ∞
5. ∞ + (-∞) = ∞ - ∞ = -∞ + ∞
6. sqrt(x) if x < 0 (if EO will not support complex arithmetics)
7. log(x) and ln(x) if x < 0 (if EO will not support complex arithmetics)
8. powr(+-0, +-0) (powr(x, y) is pow(x, y), but x ∈ [0,+∞] when in pow(x, y) x ∈ [-∞, +∞])

[Graur]

@maxonfjvipon What's up with this?