BibClassify: regression tests to accept old pdftotext

• Sometimes (e.g. on Debian Squeeze) the installed pdftotext version might not generate ideal text documents, and this might lead to not perfect results in BibClassify. For this reason the BibClassify regression tests may be failing. If this happens, we would retry the the with amended output that would suite still-good but less-ideal output generated with older pdftotext versions such as on Debian Squeeze. And the BibClassify regression tests would fail only when even these amended tests fail.