charts/opentelemetry-collector: fix CPU limit

While "256m" is a valid setting, the intention was likely to allow
4 pods to be scheduled on a single CPU, which we will currently
overshoot by 24/1000, since there are only 1000 units in a CPU
fraction.

Adjust to "250m", which would allow exactly 4 pods to share a CPU.

See e.g. https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#meaning-of-cpu

charts/opentelemetry-collector/UPGRADING.md

@@ -1,7 +1,7 @@
 # Upgrade guidelines
 
 These upgrade guidelines only contain instructions for version upgrades which require manual modifications on the user's side.
-If the version you want to upgrade to is not listed here, then there is nothing to do for you. 
+If the version you want to upgrade to is not listed here, then there is nothing to do for you.
 Just upgrade and enjoy.
 
 ## 0.62 to 0.63
@@ -47,7 +47,8 @@ As of v0.54.0 Collector chart, the default resource limits are removed. If you w
 ```
 resources:
   limits:
-    cpu: 256m
+    # CPU units are in fractions of 1000; memory in powers of 2
+    cpu: 250m
     memory: 512Mi
 ```
 

charts/opentelemetry-collector/values.yaml

@@ -270,7 +270,8 @@ ports:
 resources: {}
 # resources:
 #   limits:
-#     cpu: 256m
+#     # CPU units are in fractions of 1000; memory in powers of 2
+#     cpu: 250m
 #     memory: 512Mi
 
 podAnnotations: {}