feat: support calculate custom gpu resource

[Abirdcfly]

Why are these changes needed?

• intel use some gpu resource key like gpu.intel.com/i915
• some gpu scheduler like HAMi support set custom gpu resource key

The above two GPU settings cannot be counted in the desiredGPU field of the raycluster status.

So I added an environment variable CUSTOM_GPU_ACCELERATOR, using commas to separate the custom gpu resource key that needs to be counted.

Related issue number

Checks

• I've made sure the tests are passing.
• Testing Strategy
  • Unit tests
  • Manual tests
  • This PR is not tested :(

[andrewsykim]

#2425 added better support for custom accelerators, can you update the list of custom accelerators instead?

https://github.com/ray-project/kuberay/blob/master/ray-operator/controllers/ray/common/pod.go#L41-L43

This won't update the desiredGPU coloumn, but I think that's fine, we can't have coloumns for every accelerator