Fix for: Merging two resultsets marks uncovered line as irrelevant

[michelgrootjans]

Fixes issue #517

[GlennVB]

This might be a duplicate of #513

[hanazuki]

Only this example behaves differently against master:

  1) merge helpers SimpleCov::ArrayMergeHelper numbers and empty should eq [0]
     Failure/Error: it { expect(merge([], [0])).to eq [0] }

       expected: [0]
            got: [nil]

---

When I wrote the patch #513, I thought there might be three kinds of coverage values for a file.
Give a file having 4 physical lines, in a resultset there may be a entry for the file:

1. [0, 1, 2, nil] -- loaded; Integers for executable lines and nils for non-executable lines
2. [0, 0, 0, 0] -- not loaded but tracked by simplecov; every line is marked as 0
3. nil -- not loaded nor tracked

So I have assumed that if a file appears in two resultsets, there two coverage values are both Arrays and have the same length. In the test case above, this assumption does not hold ([] and [0] have different lengths).

If the assumption is not correct, patching as in this p-r is necessary.

[michelgrootjans]

Your assumption is correct. Both Arrays should have the same length in this version of master. I my fix one of both could be empty. In short, my PR has just become a refactor of the previous fix.
See comparison here: https://github.com/colszowka/simplecov/compare/master...michelgrootjans:master?diff=split&name=master#diff-75e6f92edee607d158835c2598d6a2c0.

Historical explanation:
When I wrote this PR, I wasn't aware of issue #513 and its fix, as we were using simplecov 0.11 with this implementation. I wrote a few specs to zoom in on the problem, and fixed it at the level that seemed appropriate.

[PragTob]

Hey there thanks for the PR!

@michelgrootjans do I understand your comment correctly that this PR isn't needed anymore?

Otherwise if I can dig up some time I'll investigate the initial case again.

[michelgrootjans]

Indeed, I'll close this PR