CI: Run a CI pipeline on a dedicated machine

This is exploratory work for %(OKR 2022Q3 - 4.3) - Faster CI. The goal is to establish whether migrating to dedicated runners would be a good idea. To do that, we need to know how long would pipelines take to run on such machines.

Suggested tasks:

• spawn a dedicated machine on AWS with good hardware;
• add it as a runner for tezos/tezos with a tag like experimental or dedicated;
• fork a branch of tezos/tezos and patch the .gitlab-ci.yml files to add this tag to all jobs;
• push the branch;
• compare the duration of each job with current runners.

Then one can also investigate:

• how many jobs we manage to run on a single dedicated machine;
• whether we can use the cache of Dune.