fix: hardcode resouce request of gpus to 1 if utilize a existing ray …

[bjwswang]

…cluster

What type of PR is this?

What this PR does / why we need it

Which issue(s) this PR fixes

Fixes #

Special notes for your reviewer

Verified on:

1. single node & 2gpus (WITH resource limit of "nvidia.com/gpu = 2")

   

2. two nodes & each have 1 gpu

   • RAY_CLUSTER_INDEX = 0
   • nvidia.com/gpu = 2

Pod logs:

[bjwswang]

@nkwangleiGIT FYI