Fix the chef config for pre not using the correct ratelimiting instances.

Currently, the pre-console nodes don't use the sentinel configuration for accessing redis ratelimiting, it connects directly to an ip configured in our chef repository and it ignores the sentinels configured there: https://gitlab.com/gitlab-com/gl-infra/chef-repo/blob/89102984ec8375d947f951f4980a607d50b95e93/roles/pre-base.json#L189

We should make sure that the config for pre is the same as it is for gstg:

• Using mymaster as the host, value gotten from gkms
• Using the sentinels to connect to the right primary.

Discovered while working on #1876 (closed)