queue: park nixl-cu13 (same source tree as nixl-cu12 PR #2314, cu12/cu13 split is CUDA-version-detection only)

Author: luhenry

.queue.yml

@@ -4980,11 +4980,11 @@ packages:
     links libglib-2.0/libgraphite2, which auditwheel would vendor and gpl_sources would then have to archive. Merged (#1819). Push-triggered build+publish on main run 35573051518 succeeded. 404 on pypi.riseproject.dev/simple/pygame-ce/ as of this check, pending docs catalog PR #2133; move to published once it merges and the index returns 200.'
 - pkg: nixl-cu13
   version: 1.4.0
-  home: ''
-  repo: ''
-  status: queued
+  home: 'https://github.com/ai-dynamo/nixl'
+  repo: 'https://github.com/ai-dynamo/nixl'
+  status: parked
   pr: null
-  notes: '10 Linux wheels upstream (abi: cp310,cp311,cp312,cp313,cp314); no riscv64 on PyPI or pypi.riseproject.dev. PyPI records no project URL; locate upstream during research.'
+  notes: 'Same ai-dynamo/nixl source tree as nixl-cu12 (PR #2314) - the cu12/cu13 split is purely which CUDA major version meson''s optional CUDA probe finds at build time, not a separate codebase. Our riscv64 build has no CUDA toolkit at all, so it can only ever produce the nixl_cu12-named wheel (see nixl-cu12''s notes for the full feasibility case: CUDA is `required: false` in meson.build, the released core .so''s link no CUDA, UCX has a riscv64 arch dir). Park this entry rather than duplicate the port; revisit only if a future release renames the split or ships a real cu13-specific build path.'
 - pkg: bsdiff4
   version: 1.2.6
   home: https://github.com/ilanschnell/bsdiff4