Fix external table/FDW metadata queries and update role-grant tests

[chenjinbao1989]

backup/queries_externals.go: gp_exttable_fdw represents an unset rejectlimittype as char(-1) ('\377'), not SQL NULL, so GetExternalTableDefinitions was leaking that sentinel through instead of coalescing it to ''.

backup/queries_functions.go: GetForeignDataWrappers and GetForeignServers had no ORDER BY, so Postgres could return rows in a nondeterministic order.

integration/metadata_globals_*_test.go: PG16 reworked role-grant semantics so an implicit grantor always resolves to the bootstrap superuser rather than the acting (super)user, and an explicit GRANTED BY now requires that role to actually hold ADMIN OPTION. Updated fixtures accordingly, and skipped "handles dropped granter" since PG16's grantor dependency tracking means a role can no longer be dropped while still recorded as another membership's grantor.

integration/postdata_create_test.go: updated a hardcoded view-definition golden string to match PG16 ruleutils, which no longer qualifies column references with the table name when only one table is in the FROM clause.

closes: #ISSUE_Number

---

Change logs

Describe your change clearly, including what problem is being solved or what document is being added or updated.

Contributor's checklist

Here are some reminders before you submit your pull request:

• Make sure that your Pull Request has a clear title and commit message. You can take the Git commit template as a reference.
• Learn the code contribution and doc contribution guides for better collaboration.
• Make sure that CICD workflow is successful.
• List your communications in the GitHub Issues or Discussions (if has or needed).
• Feel free to ask for the cloudberry committers or other people to help review and approve.

[tuhaihe]

Hi @chenjinbao1989, can these changes be used for both PG14 and PG16 based Cloudberry versions, or only for PG16 based Cloudberry?

I think we should add the main branch of Cloudberry to the CI pipeline to test these changes as well.

cloudberry-backup/.github/workflows/cloudberry-backup-ci.yml

Lines 54 to 60 in 509edb4

	env:
	CLOUDBERRY_REPO: apache/cloudberry
	# Currently targeting REL_2_STABLE (2.x) branch only.
	# TODO: Switch back to "main" once Cloudberry 3.x / main branch stabilizes with PG16 support.
	CLOUDBERRY_REF: REL_2_STABLE
	CLOUDBERRY_DIR: cloudberry
	CLOUDBERRY_BACKUP_DIR: cloudberry-backup

[chenjinbao1989]

Hi @chenjinbao1989, can these changes be used for both PG14 and PG16 based Cloudberry versions, or only for PG16 based Cloudberry?

I think we should add the main branch of Cloudberry to the CI pipeline to test these changes as well.

cloudberry-backup/.github/workflows/cloudberry-backup-ci.yml

Lines 54 to 60 in 509edb4

	env:
	CLOUDBERRY_REPO: apache/cloudberry
	# Currently targeting REL_2_STABLE (2.x) branch only.
	# TODO: Switch back to "main" once Cloudberry 3.x / main branch stabilizes with PG16 support.
	CLOUDBERRY_REF: REL_2_STABLE
	CLOUDBERRY_DIR: cloudberry
	CLOUDBERRY_BACKUP_DIR: cloudberry-backup

These changes are only for pg16 based codes.

[woblerr]

Please update the remaining PG16 ruleutils expectations in the end-to-end test:

1. key_dependent_view:
   

   cloudberry-backup/end_to_end/end_to_end_suite_test.go

   Line 2097 in 9edbb87

   Expect(contents).To(ContainSubstring("CREATE OR REPLACE VIEW public.key_dependent_view AS SELECT view_base_table.key,\n (view_base_table.data COLLATE \"C\") AS data\n FROM public.view_base_table\n GROUP BY view_base_table.key;"))

2. key_dependent_view_no_cols:
   

   cloudberry-backup/end_to_end/end_to_end_suite_test.go

   Line 2098 in 9edbb87

   Expect(contents).To(ContainSubstring("CREATE OR REPLACE VIEW public.key_dependent_view_no_cols AS SELECT\n FROM public.view_base_table\n GROUP BY view_base_table.key\n HAVING (length((view_base_table.data)::text) > 0);"))

CI is currently failing on this mismatch in Cloudberry-backup Tests (end_to_end).