Merge: Speedup Bernoulli Multi-Armed Bandit Example (#380)

This PR reduces the runtime of the Bernoulli multi-armed bandit example
from about 55 seconds to 7 seconds (in SMOKE_TEST setting) and sets a
random seed for reproducibility.

This is supposed to improve the displayed results in the example which
were unconverged and didnt show the discussed trend.

examples/Multi_Armed_Bandit/bernoulli_multi_armed_bandit.py

@@ -25,6 +25,7 @@
 )
 from baybe.surrogates import BetaBernoulliMultiArmedBanditSurrogate
 from baybe.targets import BinaryTarget
+from baybe.utils.random import set_random_seed
 
 ### An Imaginary Use Case
 
@@ -61,6 +62,10 @@ def draw_arm(self, arm_index: int) -> bool:
 
 ### Setup
 
+# Setting a random seed for reproducibility.
+
+set_random_seed(42)
+
 # For our example, we use a system with a fixed collection of win rates, which we
 # assume are unknown to us:
 
@@ -83,8 +88,8 @@ def draw_arm(self, arm_index: int) -> bool:
     qThompsonSampling(),  # Online optimization
     PosteriorStandardDeviation(),  # Active learning
 ]
-N_MC_RUNS = 2 if SMOKE_TEST else 10
-N_ITERATIONS = 2 if SMOKE_TEST else 200
+N_MC_RUNS = 3 if SMOKE_TEST else 10
+N_ITERATIONS = 50 if SMOKE_TEST else 200
 
 
 ### Building the Model