#6654 Accelerator device class Pull request opened by jeevb ## Why are the changes needed? Currently, only accelerators that use the

<http://nvidia.com/gpu|nvidia.com/gpu>

resource name in container specs are supported. This PR adds support for additional accelerator types (e.g. TPU, Neuron, AMD, Gaudi) ## What changes were proposed in this pull request? IDL only change that allows passing through the accelerator's device class from SDK. ## How was this patch tested? Tested on Union Cloud with TRN1 instances. flyteorg/flyte