Decrease probability of stochastic failures in test-slow
Our stochastic tests are supposed to fail around 1 in 100 runs. But when I'm doing a backport to 0.2.9, there are up to 14 jobs times 9 branches, each of which runs a test instance.
So let's decrease the probability to about 1 in (100 * 14 * 9).
Here's what the output looks like:
slow/prob_distr/stochastic_uniform: [forking] fail uniform sampler
FAIL src/test/test_prob_distr.c:1209: assert(ok)
NOTE: This is a stochastic test, and we expect it to fail from
time to time, with some low probability. If you see it fail more
than one trial in 100, though, please tell us.
Seed: 5DB9A3B32C29B76D7A0032700DD142BB
[stochastic_uniform FAILED]