SGE setting -V breaks setting metadata on the cluster

I've seen this error reported every so often on the mailing list, but couldn't find an issue on it:

> An error occurred setting the metadata for this dataset
> Set it manually or retry auto-detection

Apparently a standard work around is to change ``retry_metadata_internally=true`` to ``retry_metadata_internally=false`` in your configuration.

We just saw this error specific to one user account (which was an important clue as our Galaxy submits jobs to an SGE cluster where they run as the user) on Trimmomatic and FastQC jobs. This in turn broke workflows which would have used the green outputs which were lacking metadata.

We could fix this by removing the ``-V`` setting in the user's ``~/sge_request`` file.

i.e. When submitting the compute job, Galaxy must be defensively avoiding the user's default, but when submitting the metadata job, it was not - and for this user their settings broke this.

This was on an old server running ``release_16.01``, and unfortunately I am not in a position to try this on the latest release.

No source file or test is named. Start by tracing Galaxy's SGE job-submission path and compare the compute-job and metadata-job handling of the user's ~/sge_request defaults, especially -V. Reproduce with Trimmomatic or FastQC if possible; done means metadata jobs are insulated from the user's SGE setting and workflow outputs receive metadata.

Written by the indexing model from the issue text.