Must “export” release artefacts use a format-dependent ontology IRI?

Currently, the code that produces the release artefacts in “export” formats (e.g. `ttl` or `json`) forcibly re-annotate the release artefacts before converting it to the export format:

```Make
$(ONT)-full.ttl: $(ONT)-full.owl
$(ROBOT) annotate --input $< --ontology-iri $(ONTBASE)/$@ $(ANNOTATE_ONTOLOGY_VERSION) \
convert --check false -f ttl -o $@.tmp.ttl && mv $@.tmp.ttl $@
```

This should not be needed as `$(ONT)-full.owl` should already have been fully annotated. The only consequence of this re-annotation is that both the ontology IRI and the version IRI of the `$(ONT)-full.ttl` artefacts are changed to IRIs ending with `.ttl` rather than `.owl`.

Is that _really_ wanted??? I thought it was agreed that the ontology should always be identified by its `.owl` IRI, regardless of which format it is actually written in.

Start from the Make rule producing $(ONT)-full.ttl and inspect the corresponding export-format rules and annotation step. Compare the ontology and version IRIs in the generated .ttl and .json artefacts with those in $(ONT)-full.owl; done means format conversion no longer changes the agreed ontology identifiers.

Written by the indexing model from the issue text.