fact_gathering improve linux thread counts (#82261)

don't use `//` aka `floor()` as some architectures show counts of 1.9 threads per CPU and market as 2 per CPU, `round()` will be closer to that.

changelogs/fragments/thread_counts.yml

@@ -0,0 +1,2 @@
+bugfixes:
+  - fact gathering on linux now handles thread count by using rounding vs dropping decimals, it should give slightly more accurate numbers.

lib/ansible/module_utils/facts/hardware/linux.py

@@ -257,7 +257,7 @@ class LinuxHardware(Hardware):
         if collected_facts.get('ansible_architecture') == 's390x':
             # getting sockets would require 5.7+ with CONFIG_SCHED_TOPOLOGY
             cpu_facts['processor_count'] = 1
-            cpu_facts['processor_cores'] = zp // zmt
+            cpu_facts['processor_cores'] = round(zp / zmt)
             cpu_facts['processor_threads_per_core'] = zmt
             cpu_facts['processor_vcpus'] = zp
             cpu_facts['processor_nproc'] = zp
@@ -282,9 +282,9 @@ class LinuxHardware(Hardware):
 
                 core_values = list(cores.values())
                 if core_values:
-                    cpu_facts['processor_threads_per_core'] = core_values[0] // cpu_facts['processor_cores']
+                    cpu_facts['processor_threads_per_core'] = round(core_values[0] / cpu_facts['processor_cores'])
                 else:
-                    cpu_facts['processor_threads_per_core'] = 1 // cpu_facts['processor_cores']
+                    cpu_facts['processor_threads_per_core'] = round(1 / cpu_facts['processor_cores'])
 
                 cpu_facts['processor_vcpus'] = (cpu_facts['processor_threads_per_core'] *
                                                 cpu_facts['processor_count'] * cpu_facts['processor_cores'])