[Utils] DoubleMath - implement boolean comparison operations

[vsubhuman]

DoubleMath provides #compare(double,double):int operation that does not allow nulls, cuz decision whether null is greater or lower than everything else, or should throw NPE is project-specific and is up to the library user.

But! There should be explicit comparison operations except #equals:

1. #greaterThan(Double,Double):boolean
2. #greaterThanOrEquals(Double,Double):boolean
3. #lessThan(Double,Double):boolean
4. #lessThanOrEquals(Double,Double):boolean

Rules for these methods are:

1. Two nulls are equal
2. Null is neither greater, nor lower than any other number or null.
3. Two NaNs are equal
4. NaN is neither greater nor lower than any other number or NaN.
5. Any other values are compared by DoubleMath#compare

Note: NaN comparison rules are consistent with java rules for comparison operations: ==, <, <=, >, >=. Any comparison of NaN except equality is false by definition. Though two NaNs are equal.

Note: null comparison rules are same as for NaN. Any comparison of null except equality is false by definition.

Note:

NaN comparison rule breaks consistency between these explicit methods and #compare(double,double):int, cuz latter is forced to return an integer value and uses default comparison rules for double values based on the underlying bits.

So, for example:

DoubleMath.greaterThanOrEquals(12.0, Double.NaN);   // false
DoubleMath.lessThanOrEquals(12.0, Double.NaN);      // false
DoubleMath.compare(12.0, Double.NaN);               // 1

Double.NaN by bits comparison is considered to be the smallest number (smaller than negative infinity), but it IS neither greater, nor lower than 12.0 and definitely isn't equal to it.

So any consistency isn't possible in this particular case. Default #compare should be used when you need to sort values, or anything like this. But explicit methods should be used when you need binary answer to comparing two numbers by specific comparison operation.

---

Related to: #116

Note: there should not be overloaded versions of these methods for additional EPS parameter. DoubleMath.Instance should be used in case like this.