[skip CI] New versions of kubernetes always pull the latest image, not matter what the PullPolicy is, the solution is to tag this image with a predefined tag. There could be contention here if we are launching 2 buildtree instances simultaneously but anyway, there is already a problem with the tag latest of this image...

Now copies keytab between pods and changes ownership/permissions atomically

The workflow now includes the endpoint, which includes the Kubernetes namespace that the pod is
running in.

cta-cli.sss.keytab is deprecated. CTA Frontend now uses
/etc/cta/eos.sss.keytab, which should contain one SSS key per EOS
instance (only one in the case of CI).

Changing /etc/ctafrontend_SSS_c.keytab and /etc/ctafrontend_SSS_s.keytab to /etc/cta/cta-cli.sss.keytab so that CI and preprod have now the same configuration files.

Keep in mind thatL
The user in the ctafrontend SSS key is the EOS instance name, the rest is BS. I tried to make it clear in the configuration files and names when a field is useless.
WFE scripts are now the same between preprod and CI and the ctafrontend
configuration file can move in the rpm.

Using /etc/cta/cta-frontend.krb5.keytab file as ctafrontend kerberos keytab instead of /etc/cta-frontend.krb5.keytab

their namespace.
According to
https://kubernetes.io/docs/concepts/configuration/secret/#creating-a-secret-using-kubectl-create-secret:
```
Secret API objects reside in a namespace. They can only be referenced by
pods in that same namespace.
```

added a /CANSTART file so that eos waits for all keys to be OK before starting.
/bin/sh -> /bin/bash to allow function names with `-`...
yum and yum-config-manager overridden when in BUILDTREE environment
ctafrontend not booting needs further work

and generating all the required krb5 keytabs

Only eosusers and powerusers can write to it, and powerusers only can drop replicas for all file in this directory.
poweruser1 now drops the disk replica of the file before it is recalled in `client_ar.sh` test script.

plus renaming users as stated: users, powerusers, ctaadmins, eosadmins are the one we need in real life.

and create all the needed users accounts and groups on the mgm so that we can `xrdcp` files from the `client` pod.

Adding client pod with cta cli and xroot-client using `user1` kerberos ticket for authentication

2 tapeservers are now available: tpsrv01 and tpsrv02.
systest.sh only put tpsrv01 up and perform the basic test on 1 drive only.

Fixed missing dependencies in image.

Fixed directory name in docker which did not match the top one.
Split the image create to allow fast script fixing.

Default type is mhvtl

Adding the `die` function to echo and exit in case of failure as the previous `(echo "MSG"; exit 1)` only exited from the subprocess in `()`.

Adding selectors in persistent volume claim as we are now using kubernetes >1.3 and we need to manage more resources with pv than before (logs issue, others?)

Adding more explicit errors (TIMED OUT) and validating steps during instance creation. Add  option to  so that the namespace can be kept after the system test.

Using lower case only for configmap keys as Eric had a problem with upper case with the same kubectl binary... Will be fine for everyone this way.