merge/irr: harden parsing and deterministic output

[SuvidhJ]

Small hardening follow-up from regression checks.

Main changes:

1. fix IRR entry splitting for blank-line handling and final-entry flush
2. make merge output writing deterministic and avoid blank-line artifacts
3. cap merge worker count for safer behavior on constrained systems
4. add deterministic merge regression coverage

Validation:
-> pytest tests/test_merge.py tests/test_irr_parse.py -q
-> pytest -q

Issue linkage:

• addresses Harden IRR entry splitting and deterministic merge output under constrained resources #138

[Copilot]

Pull request overview

Hardens IRR parsing and merge behavior to better handle edge-case inputs and produce stable merge outputs under constrained resources (addresses #134).

Changes:

• Fix IRR RPSL object splitting to handle blank lines robustly and flush the final object without requiring a trailing blank line.
• Make merge output writing more deterministic and avoid blank-line artifacts; cap merge worker concurrency.
• Add a regression test asserting deterministic merge output.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
tests/test_merge.py	Adds a regression test to ensure general_merge produces identical output on repeated runs.
kartograf/merge.py	Caps ProcessPoolExecutor workers, adjusts chunk sizing accordingly, and rewrites output emission to avoid blank-line artifacts and improve determinism.
kartograf/irr/parse.py	Improves IRR entry splitting logic for blank-line handling and ensures the final entry is flushed even without a trailing blank line.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[fjahr]

Please try to explain what the problem(s) are that you are solving in the PR rather than describing only the change. You should also make a separate commit for each fix, ideally with a test that did not pass before the fix but passes afterwards.

Validation:

Please also test consistency with a full reproduction run from asmap-data, thanks!