Fix number of CPUs for AdvancedTreeSearch (#448)

Specifying `cpu=2` lead the job to still only consume 100% (un-normalized, so only one core) of CPU, because `OMP_NUM_THREADS=1`.

Removing the division by two leads the job to consume the desired 200%, i.e. 2 cores.

recognition/advanced_tree_search.py

@@ -239,7 +239,7 @@ def create_files(self):
         self.feature_flow.write_to_file("feature.flow")
         util.write_paths_to_file(self.out_lattice_bundle, self.out_single_lattice_caches.values())
         extra_code = "export OMP_NUM_THREADS={0}\nexport TF_DEVICE='{1}'".format(
-            math.ceil(self.cpu / 2), "gpu" if self.use_gpu else "cpu"
+            math.ceil(self.cpu), "gpu" if self.use_gpu else "cpu"
         )
         self.write_run_script(self.exe, "recognition.config", extra_code=extra_code)