Regression suite pipe8.ecl test liable to fail on multi node or multi pod system.

pipe8.ecl creates temporary files via PIPE on the local file system (in curent working directory, which will generall be /var/lib/HPCCSystems).
In runs several graphs, and it's quite possible some of the graphs will run on different Thor instances.
e.g. the graph that creates the temp files may run on Thor instance A, and the graph(s) that try to read them may run on Thor B.
If that happens, the job will fail because the parts do not exist in the local filesystem cwd on Thor B.

The test is invalid, it should create the files in shared file system space.

NB: This was noticed to fail (sometimes) whilst running on K8s with multiThorLinger set.
Meaning Thor instances could run graphs for any job, and as a consequence, multiple Thor's can be in play for the same job.
But this issue could have equally been exposed by running this test on a multi instance load-balanced Thor on bare-metal