Output orthoXML violates the OrthoXML 0.5 schema #116

Problem

FastOMA_HOGs.orthoxml does not validate against the OrthoXML 0.5 schema (tests/data/orthoxml-0.5.xsd):

1. Single-child root wrapper: when a rootHOG's top-level element is a paralogGroup (duplication at the root level), read_infer_xml_rhog wraps it in an orthologGroup with that paralogGroup as the only child (_infer_subhog.py:112-122). The schema requires at least two children.
2. Wrapper TaxRange is not the MRCA: that wrapper uses the HOG's taxonomic level as TaxRange (_infer_subhog.py:120), while HOG.to_orthoxml uses the MRCA of the members' species. The two disagree when the members span less than the level. Its scores are computed against that level as well.
3. gene@main_isoform is not an allowed attribute in OrthoXML 0.5 (collect_subhogs.py:208-209); runs with splice files produce invalid files.

Tests (strict xfail on PR #112)

• tests/test_level_processor.py::test_read_infer_xml_rhog_groups_have_two_children
• tests/test_level_processor.py::test_read_infer_xml_rhog_wrapper_taxrange_is_mrca
• tests/test_subhog_stage.py::test_orthoxml_schema[D0625237]
• tests/test_subhog_stage.py::test_taxrange_is_mrca_of_member_species[D0625237]
• tests/test_collect_subhogs.py::TestWriteHogOrthoxml::test_root_level_duplication_is_schema_valid
• tests/test_collect_subhogs.py::TestWriteHogOrthoxml::test_main_isoform_output_is_schema_valid

Proposed fix

1./2. A paralogGroup at the root level can stay as it is in <groups>, or each of its children becomes its own top-level HOG. The OrthoXML version that FastOMA claims to write determines which one is valid, so this needs a decision (the orthoxml_v03 flag suggests the wrapper was added for v0.3).
3. Write isoform information in a schema-conformant way, e.g. as a <property> or in the notes, or extend the schema upstream (qfo/orthoxml).