Right now the peak_file field of SpectrumDataset is only the base name of the origin peak file for each spectrum. This makes it difficult to track spectra that originate from peak files that have the same file name but different paths (e.g. foo/spectra.mgf and bar/spectra.mgf).