IntervalListToolsTest.testScatterByContent and IntervalListToolsTest.testScatter could use some improvement

[kachulis]

The tests IntervalListToolsTest.testScatterByContent and IntervalListToolsTest.testScatter are a bit brittle in that they make an assumption that the transformation between scatter count and scatter content is bijective, which it is not. I believe that the previous test data has only passed due to using examples where the reversibility is preserved. However, it is relatively easy to construct reasonable test data where this bijectivity is broken and one or the other of the tests will fail. This led to a bit of a messy work around in #1786, but would be better to improve the way the tests are written to avoid this pitfall.