Fix DefaultColumnWidth

[Bykiev]

This PR fixes DefaultColumnWidth type and implementation. Due to POI bug it returned baseColumnWidth, which is integer. This is a breaking change.

Fixes #943

[Bykiev]

This seems to be incorrect: when retrieving column width we should check first if the column width is set, otherwise if defaultColWidthspecified, if not, we should check baseColWidth or fallback to 8.43 otherwise

[Bykiev]

Failing test doesn't related to changes, seems #1132 has the same issue with parallel test execution.
@lahma, @tonyqus, do you have any ideas how to fix it or maybe how to run IO tests sync?

[lahma]

Would it just make more sense in the test case (TestOpenFileThenSaveDelete) to check that neither of the assigned test files exist? Honestly I don't fully understand the test case.

Assert.False(File.Exists(tempFile));
Assert.False(File.Exists(tempFile2));

[Bykiev]

Would it just make more sense in the test case (TestOpenFileThenSaveDelete) to check that neither of the assigned test files exist? Honestly I don't fully understand the test case.

Assert.False(File.Exists(tempFile));
Assert.False(File.Exists(tempFile2));

Yes, this test can be rewritten, but then another IO-dependent test can fail. We need to resolve it globally

[lahma]

I wouldn't trust the RunSerialyAndSweepTmpFilesAttribute at all, it's implementation looks quite fishy and expects that process-wide locks would work for test runners running multiple assemblies concurrently (they don't). NUnit does not run tests concurrently by default so there's that too.

I'd say removing the attribute and adjusting the test case to check for specific items would be the way to go. Some cleanup logic could go somewhere like assembly fixture teardown.

[tonyqus]

LGTM