es_ES isbn13 returns ISBN with an empty registrant segment

[feiiiiii5]

What does this change

Removes the bogus 978/13 registrant-group rules from the es_ES ISBN provider so Faker("es_ES").isbn13() no longer emits ISBNs with an empty registrant segment.

What was wrong

faker/providers/isbn/es_ES/__init__.py lists four ranges for group "13" with a registrant length of 0 (0100000-5999999, 6050000-6999999, 7350000-8749999, 9000000-9899999). When one of those ranges is drawn, _registrant_publication splits the registrant as the empty string, and ISBN13.format() joins it back as 978-13--<digits>-<check> — one empty hyphen-delimited field in roughly 44% of 2000 seeded draws on master 1879c986 (878/2000), confirmed for both es_ES and es (which inherits it) with 0/2000 for every other locale.

The malformed samples are not parseable ISBN-13: the segment between the group and the publication is empty. Every other es path (978/84) is unaffected.

How this fixes it

Drops the 978/13 group from the rules and keeps group 84 only, so every ISBN emitted is structurally well-formed. The 978/13 group with a zero-length registrant is not a usable ISBN shape, so removing it could only have produced the malformed output in question.

Fixes #2472.

AI Assistance Disclosure (REQUIRED)

• If AI tools were used, I have disclosed which ones, and fully reviewed and verified their output.

Checklist

• I have read the documentation about CONTRIBUTING
• I have read the documentation about Coding style
• I have run make lint

Test plan

pytest tests/providers/test_isbn.py
  without the fix   1 failed, 6 passed
  with the fix      7 passed

[feiiiiii5]

Closing in favour of #2476, opened ~45 min earlier, which covers the same defect.

I also had the premise wrong, and @harmehak0173 corrected it: I treated 978-13 as a Spanish group, but it does not exist in the ISBN standard. Spain's ranges are 978-84 and 979-13, so those four zero-length rules are an invalid group rather than a slip inside a valid one. Their fix removes it and also hardens format() and _check_digit().

Thanks for picking it up, and for checking the agency's RangeMessage.xml rather than taking the report at face value.