Allow the GitLab application to know what fleet it's running on to reduce cardinality of gitlab_sli metrics.

This came up as part of #1099 (closed): we want to initialize all metrics for rails endpoints. But not all endpoints are served by all kinds of fleets.

We could keep cardinality down if we only initialized the endpoints being served by the fleet that the application is starting on.

To do this, we could set an GITLAB_METRICS_INITIALIZE env variable per type.

If it's web, we only initialize with labels for Rails-controllers, if it's api we only initialize for labels for grape endpoints and GraphQL labels, if it's for git we only initialize the endpoints that are hit on the git service.

These are the endpoints we need to initialize for git: https://thanos-query.ops.gitlab.net/graph?g0.expr=sum%20by%20(endpoint_id)(sli_aggregations%3Agitlab_sli_rails_request_total_rate6h%7Btype%3D%22git%22%2C%20env%3D%22gprd%22%7D%20%3E%200)&g0.tab=1&g0.stacked=0&g0.range_input=1h&g0.max_source_resolution=0s&g0.deduplicate=1&g0.partial_response=0&g0.store_matches=%5B%5D

If the env-var is not set, we will initialize all metrics, like we currently do.