Use go mod with ctr runtime v0.2.0

[skonto]

Tested with:

GOPATH=path/go/new_runtime # required for the codegenerator to be picked up.
cd path/go/new_runtime/spark-on-k8s-operator # containts the checked out project no need for src/blablabla
export GO111MODULE=on
go fmt ./...
go vet ./...
go test ./...
go generate && CGO_ENABLED=0 GOOS=linux go build -o ~/operator
# build image
docker build --build-arg SPARK_IMAGE=lightbend/spark:2.1.1-OpenShift-2.4.3-2.12  -t skonto/operator:test -f Dockerfile .
# deploy
helm install ./fdp-sparkoperator \
  --name spark-operator-ns \
  --namespace spark \
  --version 0.4.0 \
  --set sparkJobNamespace=spark \
  --set imagePullPolicy=Always \
  --set operatorImageName=skonto/operator \
  --set operatorVersion=test \
  --set logLevel=4 \
  --debug
# run a spark app and verify things work as expected

Also tested go mod tidy required by kubebuilder cmd. WARN: this command will mess the deps.
Right now we have master versions k8s.io/* deps but we need to freeze to older ones eg kubernetes-1.14.1 as it is done upstream.
@yuchaoran2011 fyi the new branch is controller_runtime_0.2.0 and this PR starts with the latest commit upstream.
After I merge this I will continue with the webhook stuff and then apply the work you have done.

[skonto]

@yuchaoran2011 not sure if you faced this but if we move to 1.13 (eg. golang:1.13.0-alpine) we need to run:

...
WORKDIR /go/pkg/mod/k8s.io/[email protected]
RUN go mod init
...

in 1.12.x the default behavior is to succeed when you have no go.mod file but not with 1.13, it will not create one for you.
I noticed that in 1.12.5 the following and then found the related issue:

go: creating new go.mod: module k8s.io/code-generator
go: copying requirements from Godeps/Godeps.json
...

Related discussion here: golang/go#29433 and PR at the very end.