Backfill missing epic_issues rows from parent links

What does this MR do and why?

Some work_item_parent_links rows for issues under legacy-synced epics have no matching legacy epic_issues row.

• Guard against missing legacy epic_issues row in... (!249710 - merged) made the REST API tolerate the gap
• [Fix] Prevent parent links creation without a l... (!250295 - merged) fixed the write paths that created it.

This MR repairs the ~225 rows that already exist on GitLab.com.

Same repair pattern as FixSyncedEpicWorkItemParentLinks (!156657 (merged)), which fixed this inconsistency in the other direction when epic_issues was still the source of truth.

Query plan

Plan: https://console.postgres.ai/gitlab/projects/gitlab-production-main/sessions/56436/commands/160627

Safety

• Insert-only, no UPDATE or DELETE; rows that disagree with their parent link are left as-is.
• Idempotent: NOT EXISTS anti-joins plus ON CONFLICT DO NOTHING against the unique indexes on issue_id and work_item_parent_link_id, so retries and races with the live write path are no-ops.
• The write paths that created these rows are fixed as of !250295 (merged), so the population only shrinks.

How to validate locally

1. On a GDK with an Ultimate license, create a group-level epic and a project issue, then link them without the legacy row:

   link = WorkItems::ParentLink.new(work_item: WorkItem.find(issue.id), work_item_parent: epic.work_item)
   link.work_item_syncing = true
   link.save!

2. Confirm EpicIssue.find_by(issue_id: issue.id) is nil.

3. Run the migration job and confirm the row is created with the epic, the link id and the issue's namespace_id:

   Gitlab::BackgroundMigration::BackfillMissingEpicIssuesFromParentLinks.new(
     start_cursor: [Epic.minimum(:id)], end_cursor: [Epic.maximum(:id)],
     batch_table: :epics, batch_column: :id, sub_batch_size: 100, pause_ms: 0,
     connection: ApplicationRecord.connection
   ).perform

Related to

Resolves the backfill follow-up from https://gitlab.com/gitlab-org/gitlab/-/issues/612058.