fix: add W3C/OGC standard prefixes to linked_data curated context

[jdsika]

Summary

Continuation of PR #71 / issue #70 — adds 5 more W3C/OGC standard semantic-web prefixes to linked_data.curated.yaml to prevent bioregistry.upper from overriding them with uppercase or non-standard names in the merged context.

Problem

When the merged context is built (merge order: obo → go → linked_data → bioregistry.upper → prefixcc), several widely-used W3C/OGC semantic-web prefixes get incorrect canonical names because bioregistry.upper either:

1. Uppercases them (ODRL, TIME, WGS84) — problematic because rdflib prefix bindings are case-sensitive
2. Renames them (dctypes instead of dcmitype)
3. Omits them entirely (xml)

PR #71 fixed rdf and rdfs using the established pattern of adding them to linked_data.curated.yaml. This PR extends the same approach to the remaining 5 affected prefixes.

What this PR changes

New entries in linked_data.curated.yaml / linked_data.csv:

Prefix	Namespace	Without this fix	With this fix
odrl	http://www.w3.org/ns/odrl/2/	ODRL (uppercase)	odrl ✓
time	http://www.w3.org/2006/time#	TIME (uppercase)	time ✓
dcmitype	http://purl.org/dc/dcmitype/	dctypes (wrong name)	dcmitype ✓
wgs	http://www.w3.org/2003/01/geo/wgs84_pos#	WGS84 (uppercase)	wgs ✓
xml	http://www.w3.org/XML/1998/namespace	(missing entirely)	xml ✓

New regression tests:

• test_w3c_semweb_prefixes_in_linked_data — verifies the 5 prefixes resolve correctly in the linked_data context
• test_w3c_semweb_prefixes_in_dyn_merged — verifies they survive dynamic merge (load_multi_context) with correct lowercase canonical forms

Evidence for canonical lowercase forms

Each prefix's canonical form is documented by its defining W3C/OGC specification:

• odrl — W3C ODRL Vocabulary §2 uses @prefix odrl: <http://www.w3.org/ns/odrl/2/> throughout the spec. Bioregistry itself stores preferred_prefix: odrl (lowercase).
• time — W3C OWL-Time §2 states: "The suggested prefix for the OWL-Time namespace is time."
• dcmitype — rdflib 7.x ships dcmitype as its built-in prefix for http://purl.org/dc/dcmitype/. prefix.cc/dcmitype confirms this. Bioregistry uses the non-standard name dctypes.
• wgs — The W3C Basic Geo Vocabulary uses xmlns:geo="http://www.w3.org/2003/01/geo/wgs84_pos#" in its own examples. rdflib 7.x ships wgs as the built-in prefix.
• xml — http://www.w3.org/XML/1998/namespace is one of only two namespaces reserved by the XML specification itself. rdflib 7.x binds it as xml by default. It was absent from all prefixmaps contexts.

Cross-reference with rdflib 7.x:

All 5 prefixes match the default bindings from rdflib.Graph().namespaces() (rdflib 7.6.0). Of the 29 rdflib built-in prefixes, 21 already matched the merged context. After PR #71 (rdf, rdfs) and this PR (+5), only 1 discrepancy remains: geo (see below).

Known limitation: geo (GeoSPARQL)

The OGC GeoSPARQL 1.1 spec states "The suggested prefix for this namespace is geo" (for http://www.opengis.net/ont/geosparql#), and rdflib 7.x agrees. However, this cannot be fixed via linked_data because the obo context (which has higher merge priority) already claims GEO for Gene Expression Omnibus. This is a known architectural conflict documented in EXPECTED_OBO in the test suite.

Action needed from maintainers

merged.csv is not regenerated in this commit (would require running the full ETL pipeline with bioregistry dependency). Please run make etl or trigger the refresh workflow to propagate the linked_data changes into the merged context CSV.

Closes #70

[matentzn]

I personally like all of these changes. I will assign someone from the core team to contemplate consequences.

BTW, no action needed but generally this is where we try to consolidate prefixes: https://semantic.farm/