fixed CollectionShouldOnlyHaveUniqueItemsByComparer gives invalid code

[Meir017]

closes #63

[Meir017]

@jnyrup with this fix the CollectionShouldNotBeNullOrEmptyAnalyzer will not report statements that have a method call between the NotBeNull & NotBeEmpty

var actual = new[] { "text1", "text2" };
actual
  .Should().NotBeNull()
  .And.HaveCount(2)
  .And.NotBeEmpty();

[jnyrup]

I guess one can always write arbitrarily obfuscated code that will hinder an analyzer in detecting patterns.

E.g. this valid but weird code

list.Should().NotBeEmpty().And.Should().NotBeNull();

is with 0.11.2 transformed into the uncompilable:

list.Should().NotBeNullOrEmpty().And.NotBeNull;

I have previously blindly trusted analyzers as "knowing the truth" and just applied its suggestions to e.g. an entire project, only to later find out that the applied code fixes had semantically changes, that broke the code.
I guess for somebody false positives could be a blocker for their adaption of the analyzer.
Roslynator e.g. chose to remove some analyzers, as their correctness could not be guaranteed and hence to could lead to invalid code.

[Meir017]

this case is related to the same issue that the analyzers used to look for the method in sequence but allowed method calls to exists between them, now they have to be one after another

[jnyrup]

And just to answer your question: I think it's fine to sacrifice functionality over correctness.