Bump xbrlkit to 0.7.3

[jfrench9]

0.7.3 fixes a term collision in the holon's canonical context. 0.7.0 added a second "label" key, which silently shadowed the "label": "rdfs:label" already there — so every holon xbrlkit wrote in 0.7.0–0.7.2 bound label to rs:label instead. This platform's own holon context binds it to rdfs:label, so the two disagreed about one term across two documents that both call themselves holons.

Why the floor moves, not just the lock

>=0.7.2 would still resolve a version that writes the wrong binding, and the SEC artifact path (adapters/sec/processors/artifacts.py) goes through xbrlkit's holon writer. 0.7.3 is the minimum that produces correct output, not merely the newest — so it belongs in pyproject.toml, not only in the lock.

Nothing changes today

• RoboLedger's holon comes from the platform's own encoder (operations/serialization/rdf/), untouched by any xbrlkit version
• its Tavi comes from an emitter byte-identical since 0.6.0
• the SEC download schedule is DefaultScheduleStatus.STOPPED, so no artifact is being written on either version

This is the prerequisite for regenerating the SEC holons on the CDN, which waits on the serialization waist's phase 3 so the vocabulary is published once rather than twice.

Verification

• 2,513 tests pass across operations/serialization, adapters/sec, schemas and operations/roboledger, including test_xbrlkit_parity.py
• just test-code clean
• the lock moves xbrlkit alone: 0.7.2 -> 0.7.3, no other churn

🤖 Generated with Claude Code

https://claude.ai/code/session_01EiGTpXmxwZjxzaRoVs4VxB

[github-actions]

Change classification: standard — only dependency manifests, lockfiles, documentation, release notes or generated output changed.

Under the change standard a standard change is gated by the required test check alone; no secondary review is required. Recorded by the change-classification job.