Enhance nullability check for "==" and "!=" operators for LINQ provider

[maca88]

This PR adds a slightly complex logic that determines if it should add a OR <Property> IS NULL check for == and != operators, fixes #1860. The logic will not add the additional null check for the following scenarios:

• The property is a value type.
  Example: s.Query<Order>().Where(o => o.Number != 5)
• The property is mapped as not null.
  Example: s.Query<Order>().Where(o => o.Status != "NEW")
• The property has a not null check before or after the != operator.
  Example: s.Query<Order>().Where(o => o.Status != null && o.Status != "NEW")
• The nullable property has a HasValue check before or after the != operator.
  Example: s.Query<Order>().Where(o => o.Number.HasValue && o.Number != 3)
• The expression for the != operator is not null.
  Example: s.Query<Order>().Where(o => (o.NotNullProperty + o.NotNullProperty2) != "TEST")
• Two nullable properties are compared with == operator and one of them or both have a not null check.
  Example: s.Query<Order>().Where(o => o.NullProperty != null && o.NullProperty == o.NullProperty2)

There are some scenarios where the logic will add a null check where currently it is not added:

• Using a subquery with an aggregate method that can return null when the subquery has no records.
  Example: s.Query<Order>().Where(o => o.Rows.Max(r => r.Number) != 0)
• Using a property that is translated to a sql function on a property that can be null.
  Example: s.Query<Order>().Where(o => o.Status.Length != 3)

In addition, the Equals method was modified to use the same logic as the == operator.

[maca88]

Changed to o.Output + o.Output as otherwise it fails in oracle as in oracle concatenating null with a non empty string will not result in a null value like in the other databases. I think that we should add an additional logic when concatenating strings in order to have a consistent behaviour across various databases and possible align with Linq-to-Object so that null + "test" will not result in a null value.

[maca88]

Removed the Take call as it produces an invalid sql statement for Oracle.

[maca88]

I've had to use lazy as otherwise an invalid query is generated due to the missing table alias for the column used in the formula.

[maca88]

Copied from #2079. It is a fix for including nested components.

[maca88]

I had to add lazy="false", otherwise some tests in QueryReuseTests (CanReuseAfterFirst, CanReuseAfterFirstOrDefault, CanReuseAfterSingle, CanReuseAfterSingleOfDefault) would fail.

[maca88]

Rebased and fixed TryGetEntityName method to work with custom entity names. I am not sure what to do with the current CodeFactor issues, I think that we should allow those issues for the test project and increase the allowed complexity a little bit :)

[hazzik]

I am not sure what to do with the current CodeFactor issues, I think that we should allow those issues for the test project and increase the allowed complexity a little bit :)

I will agree with CodeFactor that this method is too complex (and too hard for me to understand).

[maca88]

Ok, I will try to make the method more readable by adding documentation, split into multiple methods, try to simplify the logic and add isolated tests.

[maca88]

I tried my best to make the method more readable and to reduce its complexity in #2036. I've added several isolated tests in order to cover all kind of mappings. I think that the method is now more readable but the complexity didn't dropped, it even increased so I am not sure what to do about it. Should I split it into more smaller methods or replace the method with a method object?

EDIT:
By splitting the method into two, solved the CodeFactor issue.

[maca88]

Rebased and updated code to use ExpressionHelper.TryGetMappedNullability.