add env var for single-dsid mode skimming

In order to run CAF based skimming on the grid we need a more static way to construct the output names of skimmed xAODs. As panda only sends single files (with names essentially announced at submission time) to an output stream without enclosing them in a tgz we should offer an env var that, when set, overrides the auto-generated output names. The inclusion of AMI tags and DSID in the file name is then to be done via the --addNthFieldOfInDSToLFN option of panda (and suitably named input datasets which also need to be single-DSID datasets!).

When this fixed output name is used the skimming algorithm should also enforce that no more than a single DSID is encountered and make the execution fail if a second DSID is encountered (and thus making the job fail!).