rdo2: BM jobs failing b/c concurrent pip installs are failing due to sharing pip cache

[Matt Young]

We appear to be hitting instances of https://github.com/pypa/pip/issues/5345

---

in RDO Phase 2 (all branches) jobs. The jenkins configuration is using 10 executors on the BM slave node. The job creates the env, and when pip install happens some cache funk occurs and we end up with missing packages..

16:13:23 No matching distribution found for appdirs>=1.3.0 (from openstacksdk>=0.13.0->os-client-config->-r requirements.txt (line 5))
16:13:23 python setup.py install failed

although similar errors have been occuring with a wide variety of other packages (ansible, etc)

---

Have implemented the following on jenkins server for this node:

```
XDG_CACHE_HOME=$HOME/.cache/$EXECUTOR_NUMBER
```

this allows a discreet pip cache per executor.

[Matt Young (halcyondude)]

this appears to have worked, no commit as this is change to the jenkins master configuration on internal server.