ImpliedLosses score is wrong: species tree nodes are compared by identity #115

Problem

The ImpliedLosses score in the orthoXML is too high. A complete 1:1 family over all 30 species of fungi-30 (rootHOG F0848294, CompletenessScore 1.0 at every level) gets ImpliedLosses=22 at Fungi, where it should be 0. Roughly, every nested orthologGroup adds one loss. Families that really lost clades are also wrong: with species A and C of ((A,B),(C,D)) the expected value is 2.

Cause

_count_untouched_clades (_hog_class.py:82-86) decides with tree_node in reached_nodes / tree_node in r.get_ancestors() whether a clade is reached. ete3 TreeNode has no __eq__, so this compares object identity. The nodes come from different tree objects:

• the scoring node is taken from full_species_tree, a copy() of the species tree (_hog_class.py:576, _infer_subhog.py:118);
• each HOG's _tax_now points into the pruned working tree;
• every taxonomic level is written to and read from its own pickle file (infer_hogs_this_level), so the HOGs of different levels point into different copies of the tree.

The existing unit tests (tests/test_hog_scores.py) build everything on one tree object, so they don't catch it. CompletenessScore and TCSScore are not affected (they use names / the size feature).

Tests

• tests/test_hog_class.py::test_implied_losses_against_full_species_tree (strict xfail on PR Test suite ahead of the ete3 → ete4 migration #112; covers in-memory and pickled-per-level)
• the golden snapshots in tests/data/fungi30/golden/ record the current, wrong values and will change with the fix.

Proposed fix

Compare species tree nodes by name (names are unique after check-input), or resolve each _tax_now into the scoring tree via _species_name_index before comparing. More generally, storing the taxon name in HOG instead of a TreeNode would remove this class of bug and shrink the per-level pickles (each currently contains a copy of the species tree); that also helps the ete4 migration.

After the fix, regenerate the golden snapshots and check that the diff only changes ImpliedLosses values.