Using Rancher repo instead of harbor repo.

Change all the repository in the istio to use rancher repo to pull the
image.

Problem:
Registry pod stuck in AttchVolumeFailed status on upgrade. Deadlock happens where the new pod waits the old pod to dettach the volume, and the old one waits for the new one to be ready.

Solution:
Set maxUnavailable to 1, so that the old pod can be released before the new pod is ready.

  * add istio and certmanager CRDs
  * add rancher system default registry config
  * add rancher proxy to the Grafana, Prometheus, Kiali and Tracing
  * remove not used demo values.yaml files and serviceGraph(deprecated)

Add globalRegistryMode to enable deploying the Harbor as a global
registry in Rancher. It use the same domain as Rancher server in
ingress rules, and use the same tls-rancher-ingress secret for TLS.

Co-authored-by: Guangbo <guangbo@rancher.com>
Co-authored-by: gitlawr <lawrleegle@gmail.com>

Do Helm lint validation and Helm unittest for the charts

Problem:
alertmanager 0.16.0 unable to send email notification without SMTP authentication
Solution:
upgrade to alertmanger 0.17.0 which solve this problem

Issue:
https://github.com/rancher/rancher/issues/20060

**Problem:**
Buffering proxy data causes client to receive slow

**Solution:**
Disable buffer via `proxy_buffering  off;` to response client directly

**Issue:**
https://github.com/rancher/rancher/issues/19689

**Problem:**
Upgrade the operator would have errors because we change the version of
the operator charts.

**Solution:**
Keep the old version 0.0.2 the operator used before.

**Problem:**
There are two kubelet scraping targets on Prometheus, one is scraping `/metrics`, another one is scraping `/metrics/cadvisor`.
The metrics from `/metrics` endpoint are not including `container_name`.
So the `container_*` expression will double the actual mount without
`conatiner_name!=""`

**Solution:**
Add `container_name!=""` into the expression

**Issue:**
https://github.com/rancher/rancher/issues/20162

**Problem:**
The `Flunetd` pane from `Rancher Components` dashboard cannot show
the right counting of fluentd Pods

**Solution:**
Change `sum(kube_pod_info{pod=~"fluentd.*"})` to `sum(kube_pod_info{pod=~".*fluentd.*",pod!~".*aggregator.*"})`

**Issue:**
https://github.com/rancher/rancher/issues/19722

Fix the issue that node exporter crash when deploying
into the node without internal ip.

Problem:
update configure reloader to mirrored rancher image, logging configure
secret has precan and generted data
Solution:
update configure reloader image, separate configure secret into two
secrets

Issue:
https://github.com/rancher/rancher/issues/19836

- Use project name & cluster name info from Values.global

- Embed operator as sub charts
    + Support to configure operator like other charts
    + Adjust operator default limit
- Add permission to kube-state exporter
- Replace localhost by 127.0.0.1 on prometheus-auth
- Increase Nginx proxy buffers
- Configure PVC name of Prometheus or Alertmanager
    + Allow to configure PVC name of Prometheus or Alertmanager via `prometheus.persistence.name` or `alertmanager.persistence.name`
- Adjust Cluster Monitoring scrape logic
    + Don't scrape Monitoring namespace on `prometheus-io-scrape` job
    + The rate to scrape is using global interval, the default is 60s
    + Remove useless Prometheus record rules

**Issue:**
- https://github.com/rancher/rancher/issues/19693
- https://github.com/rancher/rancher/issues/18830
- https://github.com/rancher/rancher/issues/19243
- https://github.com/rancher/rancher/issues/19689
- https://github.com/rancher/rancher/issues/19410
- https://github.com/rancher/rancher/issues/19248

**Problem:**
When we start nginx in our start-up script, the nginx process would
become the child process of start-up script process and not process 1.
In this case, the kill signal from kubelet/docker will be sent to
start-up script instead of nginx so the nginx process won't stop after
kill.

**Solution:**
Change the proxy command and let nginx start at process 1.

In system-charts, we need to use the `repository` and `tag` to define
container's image name. After that, we can collect them together and
provide an images list we need for system charts.

problem:
before fluentd 1.3.1 version can't support add client cert for fluentd output

Solution:
upgrade fluentd to 1.3.3, but the related kafka gem also upgrade small
version, tested fluentd and kafka after upgrade version

Issue:
https://github.com/rancher/rancher/issues/18396

- Adding checksum over secrets to ensure change in secrets upgrades deployment
- Using rancher image for ensuring airgap case works too
- Adding nodeSelector to ensure the workloads never schedule to the Windows node
- Adding resource limits
- Add private image registry for airgap case

We will be keeping up with the upstream chart
https://github.com/helm/charts/tree/master/stable/external-dns

**Problem:**
Cannot start "rules-configmap-reloader" container with 10Mi limit resource

**Solution:**
Update images:

- quay.io/coreos/prometheus-operator:v0.29.0 ->
rancher/coreos-prometheus-operator:v0.29.0
- quay.io/coreos/prometheus-config-reloader ->
rancher/coreos-prometheus-config-reloader:v0.29.0
- prom/alertmanager:v0.16.1 -> rancher/prom-alertmanager:v0.16.1
- prom/prometheus:v2.7.1 -> rancher/prom-prometheus:v2.7.1
- grafana/grafana:5.4.3 -> rancher/grafana-grafana:5.4.3
- prom/node-exporter:v0.17.0 -> rancher/prom-node-exporter:v0.17.0
- quay.io/coreos/kube-state-metrics:v1.5.0 ->
rancher/coreos-kube-state-metrics:v1.5.0

**Issue:**
- https://github.com/rancher/rancher/issues/17997
- https://github.com/rancher/rancher/issues/18353

**Problem:**
Enable logging and monitoring in `rancher/rancher:master`, but can't see fluentd metric

**Solution:**
Consist label and endpoint name in `system-chart/rancher-monitoring:v0.0.2``

**Issue:**
https://github.com/rancher/rancher/issues/18327

**Patch:**
https://github.com/rancher/system-charts/pull/17