oneDNN container fixes

[ashahba]

fixes #1958

To test this PR:

• Copy the provided .bazelrc to tensorflow_serving/tools/docker
• Edit the Dockerfile.devel-mkl at line 104 and add the following line: COPY ./.bazelrc /tensorflow-serving/

This is how I ran it:

#!/usr/bin/env bash

export TF_SERVING_VERSION=2.8.0
export TF_SERVING_VERSION_GIT_BRANCH=master
export TF_SERVING_VERSION_GIT_COMMIT=HEAD

docker build \
  ${DOCKER_BUILD_ARGS} \
  --build-arg TF_SERVING_VERSION=${TF_SERVING_VERSION} \
  --build-arg TF_SERVING_BUILD_IMAGE="intel/intel-optimized-tensorflow-serving:${TF_SERVING_VERSION}-devel-mkl" \
  --build-arg TF_SERVING_BUILD_OPTIONS='--config=mkl -c opt --copt=-O3 --copt=-march=native' \
  --build-arg TF_SERVING_BAZEL_OPTIONS='--local_ram_resources=HOST_RAM*0.8 --local_cpu_resources=HOST_CPUS-4' \
  --build-arg TF_SERVING_VERSION_GIT_BRANCH=${TF_SERVING_VERSION_GIT_BRANCH} \
  --build-arg TF_SERVING_VERSION_GIT_COMMIT=${TF_SERVING_VERSION_GIT_COMMIT} \
  --build-arg BAZEL_VERSION=5.0.0 \
  -t intel/intel-optimized-tensorflow-serving:${TF_SERVING_VERSION_GIT_COMMIT}-devel-mkl-mkl \
  -f Dockerfile.devel-mkl \
  .

[ashahba]

Once the PR is merged the COPY is no longer needed and things would work as expected.