If 'NaN' then the worker balance summary will fail with a TypeError.
0 makes sense in the common cases:
* either you have worker tags and they are numbered
* or you have no worker tags and thus you have just one worker, and
we can call that worker '0' because it is the first one
The 'NaN' shows up when you try to use subunit-trace in commands
such as
python -m subunit.run discover <dir|package> |subunit-trace
which is something people are wont to do.
Change-Id: Ibe18f16c0c360fc5a995fcf84e5bf2b6c16abd9c
d588748a36