Improvements from our RNA-Seq pipeline

[arteymix]

Here are a few things we've improved in RSEM for our RNA-Seq pipeline.

• use move instead of system(mv...) as it works across filesystems
• use a temporary directory with auto cleanup for storing intermediary files, which is very friendly for HPC systems storing temporary results in scratch
• export STAR Log.final.out to the final directory
• fix incorrect examples in rsem-calculate-expression docs

I'm not expecting everything here will be retained, especially the shared memory option for STAR. I can reintroduce the two removed parameters to adjust the temporary dir location and conditionally set CLEANUP.