Add Dockerfile for installing kit as a KServe ClusterStorageContainer

[amisevsk]

Description

Add a Dockerfile to enable using kit as a ClusterStorageContainer for KServe via the following CR:

apiVersion: "serving.kserve.io/v1alpha1"
kind: ClusterStorageContainer
metadata:
  name: kitops
spec:
  container:
    name: kit-storage-initializer
    image: <kit-kserve-image>
    imagePullPolicy: Always
    env:
      - name: KIT_EXPORT_FLAGS
        value: ""       # Optional: Add extra flags for the `kit export` command
    resources:
      requests:
        memory: 100Mi
        cpu: 100m
      limits:
        memory: 1Gi     # Note: Default value, not measured
    supportedUriFormats:
    - prefix: kit://

This will allow downloading modelkits by specifying storageUris with kit:// -- for example

apiVersion: "serving.kserve.io/v1beta1"
kind: "InferenceService"
metadata:
  name: "torchserve"
spec:
  predictor:
    model:
      modelFormat:
        name: pytorch
      storageUri: kit://<modelkit-reference>

Note that KServe inference services generally assume a certain file structure for the model/configuration; as a result building dedicated modelkits for KServe is still likely required.