Set agent CPU resources requests/limits to exactly .5 CPUs

[smuth4]

While 512m is technically a valid quantity, it implies that the CPU is a binary-oriented resource like memory, instead of decimal-oriented.

I'd consider this only a patch change, since despite it being a change to the default values, I doubt there's anyone out there strictly relying on those extra 12 milli-CPUs being available. It does potentially mean that more agents will get scheduled than before in a cluster that generally uses "round" numbers for scheduling (e.g. if a node had 1 whole CPU free, it could now schedule 2 agent pods instead of just 1), but the whole point of requests/limits is to allow kubernetes to pack in pods as effectively as possible.

What does this PR do?

• Fixes #

If you modified files in the ./charts/jenkins/ directory, please also include the following:

### Submitter checklist
- [x] I bumped the "version" key in `./charts/jenkins/Chart.yaml`.
- [x] I added a new changelog entry to `./charts/jenkins/CHANGELOG.md`.
- [x] I followed the [technical requirements](https://github.com/jenkinsci/helm-charts/blob/main/CONTRIBUTING.md#technical-requirements).
- [x] I ran `.github/helm-docs.sh` from the project root.

Special notes for your reviewer