Add a retry for the kube api connection

[kayrus]

What happened:

Deploying gardener from scratch may fail because of the loadbalancer readiness:

*** sowing species gardener/virtual
Plugin: git[deploy] (/sow/plugins/git/plugin)
Cloning repo https://github.com/gardener/gardener.git (tag 0.27.6
)...
Initialized empty Git repository in /mounted/dev/gen/gardener/virtual/gardener_git/repo/.git/
Reinitialized existing Git repository in /mounted/dev/gen/gardener/virtual/gardener_git/repo/.git/
remote: Enumerating objects: 7255, done.
remote: Counting objects: 100% (7255/7255), done.
remote: Compressing objects: 100% (5575/5575), done.
remote: Total 7255 (delta 2049), reused 3976 (delta 1210), pack-reused 0
Receiving objects: 100% (7255/7255), 9.49 MiB | 8.29 MiB/s, done.
Resolving deltas: 100% (2049/2049), done.
From https://github.com/gardener/gardener
 * tag               0.27.6     -> FETCH_HEAD
Plugin: helm[deploy] (/sow/plugins/helm/plugin)
    using /mounted/dev/kubeconfig_sa
Running helm template for gardener ...
Plugin: kubectl[deploy] (/sow/plugins/kubectl/plugin)
manifest for Namespace(v1): garden
error: unable to recognize "/mounted/dev/gen/gardener/virtual/kubectl_apply/manifests/0-0-Namespace.garden.yaml": Get https://gardener:443/api?timeout=32s: net/http: TLS handshake timeout

Sometimes it takes time for the LB to be available.

What you expected to happen:

sow should show that the LB is not available, but it will retry the action.

[Diaphteiros]

The latest sow release features retry logic for failed kubectl and helm commands. By default, sow will retry a failed command up to four times after waiting for 15 seconds. This accumulates to a total of five attempts over 60 seconds.

I will close the issue since this should probably solve the problem, please reopen it if you object.