approval-voting: Make tests deterministic (#3899)

With random connectivity and latency is hard to actually figure it out a
delta in the benchmarking, so disable them in order to get full
deterministic behaviour when measuring performance.

At least on my machine with this configuration the results for
approval-throughput are really similar between subsequent runs:

```
CPU usage, seconds                     total   per block

approval-distribution                36.9025      3.6902
approval-distribution                36.7579      3.6758
approval-distribution                37.0418      3.7042
approval-distribution                37.0339      3.7034
approval-distribution                36.9342      3.6934
approval-distribution                36.7177       3.6718



approval-voting                      52.7756      5.2776
approval-voting                      52.5999      5.2600
approval-voting                      53.2158      5.3216
approval-voting                      53.2493      5.3249
approval-voting                      52.8524      5.2852
approval-voting                      52.8611      5.2861
approval-voting                      52.8210      5.2821
```

---------

Signed-off-by: Alexandru Gheorghe <alexandru.gheorghe@parity.io>