repository: stop migration_mirror_interval planning an update every run

Why:
- The go-gitea/gitea provider schema defaults migration_mirror_interval to
  "8h0m0s", but on read it writes repo.MirrorInterval straight back into state,
  and Gitea returns an empty MirrorInterval for non-mirror repositories.
- So the stored value ("") never matches the schema default ("8h0m0s"), and tofu
  plans an in-place update of every gitea_repository on every run even with no
  config change, keeping terraform-git plans perpetually dirty.
- migration_mirror_interval is the only migration_* field read back from the API;
  the rest are write-only migration options, so no sibling field churns.

Change:
- Add a targeted ignore_changes on migration_mirror_interval at the repository
  module so the provider default no longer fights Gitea's empty value, making
  every repository idempotent. It is a migration-only knob with no drift to track
  for these non-mirror repos.

modules/gitea_instance/modules/repository/main.tf

@@ -17,4 +17,9 @@ resource "gitea_repository" "this" {
   website                 = var.website
   autodetect_manual_merge = var.autodetect_manual_merge
   archive_on_destroy      = true
+
+  lifecycle {
+    # migration_mirror_interval defaults to "8h0m0s" but Gitea returns an empty MirrorInterval for non-mirror repos, so the read-back never matches and plans a spurious update every run; it is a migration-only knob with no drift to track here.
+    ignore_changes = [migration_mirror_interval]
+  }
 }