Drop some hard-coded IDs in migration tests

What does this MR do and why?

Some tests created namespaces and their traversal IDs with hard-coded primary IDs. However, these hard-coded primary key values do not increase the database sequence number. If another namespace is created without an ID, it's possible to hit a duplicate key error (such as in https://gitlab.com/gitlab-org/gitlab/-/jobs/3204595559).

To avoid this, don't hard-code the IDs and update the traversal IDs after creation.

How to set up and validate locally

Previously this would fail:

RAILS_ENV=test bin/rake db:reset
CACHE_CLASSES=true bundle exec rspec --format d --fail-fast ee/spec/lib/ee/gitlab/background_migration/backfill_epic_cache_counts_spec.rb ./ee/spec/lib/ee/gitlab/background_migration/backfill_project_statistics_container_repository_size_spec.rb

MR acceptance checklist

This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.

• I have evaluated the MR acceptance checklist for this MR.