Backport Nexus addressing and Avro requests to stable Python 2.5.5

[rmcdaniel]

Stable Python Nexus request fixes

Backport the already merged #113 and #114 corrections to the published 2.5.4 line and prepare ordinary patch 2.5.5. These fixes have only been published in 2.6 prereleases so far. This PR uses a temporary maintenance base to publish the stable request fixes independently of the pending leased workflow authoring changes.

• Encode Nexus arguments with the portable Avro codec, including runtime transport for oversized arguments.
• Address the target catalog and payload transport explicitly without mutating the caller's namespace, worker identity or credential role.
• Preserve deterministic target identity and legacy same-namespace replay, with both official codec/replay fixtures and focused tests.
• Set the exact version, product train and registry version to 2.5.5. The existing worker protocol stays 1.19.

Normal CI is dispatched at the exact branch because this temporary PR base is outside the workflow's main-only PR filter. Merge and publication require its full supported-Python, lint, replay/corpus, documentation and real HTTP checks. Publication must verify registry distributions and the actual installed stable SDK receiving intact arguments at a PHP worker in both namespaces, including oversized arguments. Keep #111/#112 open until their published acceptance checks pass. This does not qualify the separate leased authoring lifecycle or complete Nexus suite.

Delete both temporary remote branches after delivery. No existing tag moves.

[rmcdaniel]

Stable Python Nexus backport review

PR #120, candidate add87c4caccc9a84f868b9e4202a39044c85f97b, tree e15715254cfd87f9b7415fedf4f9797ac29c6b78. Base is immutable stable 2.5.4 at 86cfced48a3938b41d707f26eea1ff098e07ffcb.

The runtime and tests exactly match the already merged #113/#114 source at 6778efbc2d08307feac8d34893b09b9d3c7af4eb, verified by an empty Git diff over src/tests. The only later change sets project version, product train and registry version to 2.5.5. Leased authoring #117 and the 2.6 prerelease metadata are excluded. Worker protocol remains 1.19.

Review covers portable Avro argument framing, oversized argument transport, target-scoped discovery/storage, unchanged original worker namespace, control credential use and deterministic target identity with legacy same-namespace replay. Both codec and replay regressions are retained. Local Python 3.13 verification passes lint, all 32 source files' type checks and 541 selected tests, with two skipped. Hosted run 38160065136 passes Python 3.10/3.11/3.12, lint, official corpus/policy, exact built-package metadata and packaging smoke. Its real HTTP integration remains required before merge and publication. Documentation build and installed registry-only namespace/inline/external PHP-worker evidence remain required. Existing issue acceptance is not marked complete from source checks or prior prerelease evidence.

Both temporary maintenance branches must be removed after delivery. No existing tag moves. Evidence retained through January 9, 2027, extended while unfinished. No paid resource or persistent local runtime was created; the local Python container has exited and removed itself.

[rmcdaniel]

Cleanup complete after the published 2.5.5 verification. The isolated test containers, network, disposable dependencies, clean worktree and both merged local/remote branches are removed. Temporary images used only by this check are removed too. Useful raw verification evidence and the successful consumer fixtures are retained with a bounded retention date. Other contributors’ active resources are preserved.