Add regression test for XCTAssert* failure messages

[briancroom]

This backfills tests for the behavior introduced in #22.

It also fixes a minor formatting regression introduced in e5cf2f4 in which the output of XCTAssertNil incorrectly prints the actual value as an optional.

[modocache]

Could we reverse the conditional here, to avoid the !? Exclamation points make me nervous. 😅

How about:

if let value = expression() {
    return .ExpectedFailure("\"\(value)\"")
} else {
    return .Success
}

[briancroom]

I had initially chosen to not reorder it that way because I liked the symmetry of the conditionals between the various assertion functions, but I have no problem flipping it.

[briancroom]

I just reverted it to the original ordering at @eschaton's request. AFAIK there is no clean way to avoid the force-unwrap when doing the nil check first like this. Using guard could allow keeping the ordering the same, but I feel it be a pretty bad abuse of the expected semantics of that keyword.

Is there an alternative I'm overlooking? @modocache

[modocache]

Fine with me! I defer to him. 😄

[briancroom]

Rebased and fixed up now.

[eschaton]

I strongly prefer the if predicate to “match” the intent of the assertion; that’s why I wrote them as if _ { return .Success } else { return .ExpectedFailure(“_”) }.