Add namespace visibility + admin criteria to tier_badge experiment target population

PMs & PDI have been discussing the results of this experiment so far in the rollout issue.

We have a hypothesis that the inclusion of public namespaces in this experiment (because we targeted all free namespaces) could be one factor for why we don't observe a larger difference in trial start behaviour. Public namespaces are very unlikely to start a trial and convert to paid. Additionally we discovered that non-admins were being exposed to the experiment as well. Non-admins are not able to start trials/convert to paid.

Slack convo https://gitlab.slack.com/archives/CDLCBGEDV/p1698170277223769

New target population criteria

• Only include private namespaces
• Only include those with permissions to access to billing page