Recreating mzMLs when those are incomplete gets wrong servershare

[glormph]

This may be related to or part of #51 but I'm not sure. Anyway, when some files of a dataset are mzML but not all (e.g. mzml jobs have been aborted or new files have been added, reruns etc), then you queue a new mzml convert job: it will set all the files to servershare = result share -> mzml_in share. This is because that is done in a create_or_update job on StoredFile in the mzmlentry creator function.

• servershare wrong
• md5 updated to mzml_${rawfile_md5} ( reported by Yanbo), crashes refine later on. -> need to recalc any non-deleted mzML file's MD5
• Also, when mistakenly starting 2 convert jobs, the second will change all the paths to the dset.storage_loc, before the files have been moved there, in create_or_update function. Then move_single_file (after conversion) is no longer run! Maybe the create_or_update should be get_or_create instead?