Input validation: crashes and misleading errors in check-input and infer-roothogs #118

Problem

Several inputs make fastoma-check-input or fastoma-infer-roothogs crash or stop with an error that does not point at the cause. Grouped here because the fixes are small and independent.

Input folders

• Hidden files are taken as proteomes or hogmaps: ._AQUAE.fa (AppleDouble files that macOS tar/zip add to archives) or .DS_Store. check-input then stops with "These species are not in the species tree: ._AQUAE" (check_input.py:39,91, _utils_roothog.py:165-170). This broke the archive-input nf-test on Linux CI.
• A file without an extension (e.g. README) in the proteome folder crashes parse_proteomes (file.rsplit, _utils_roothog.py:166); check-input only warns about it.
• Dots are dropped from species names derived from hogmap file names ("".join(i.split(".")[:-2]), check_input.py:91), so Homo.sapiens.fa.hogmap never matches the proteome Homo.sapiens.fa.
• An empty proteome raises ZeroDivisionError (check_input.py:65) instead of being reported as invalid.

Species tree

• Duplicate internal node names are never reported: the name list is reset inside the loop (check_input.py:100).
• A tree without internal nodes raises UnboundLocalError (check_input.py:106).
• An internal node name made only of special characters becomes empty (check_input.py:168).
• An internal node named like a leaf makes ete3's prune fail with "Ambiguous node name" (check_input.py:180). Such names would also clash later, because per-level pickle files are named after nodes.
• read_species_tree calls sys.exit() (exit status 0) when the tree can't be read (_utils_subhog.py:78,81), so Nextflow sees the task as successful.

Splice files

• A species whose splice file has a single line is ignored (len(isoforms) > 1, check_input.py:195).

Tests (strict xfail on PR #112)

• tests/test_check_input.py::TestCheckProteomeFiles::test_hidden_files_ignored
• tests/test_check_input.py::TestCheckHogmapFiles::test_hidden_files_ignored
• tests/test_roothog_io.py::TestParseProteomes::test_hidden_files_ignored
• tests/test_roothog_io.py::TestParseProteomes::test_non_fasta_files_ignored
• tests/test_check_input.py::TestCheckHogmapFiles::test_dotted_species_name
• tests/test_check_input.py::TestCheckProteome::test_empty_proteome_rejected
• tests/test_check_input.py::TestSpeciesTreeChecks::test_internalnode_duplicates_reported
• tests/test_check_input.py::TestSpeciesTreeChecks::test_internalnode_single_leaf
• tests/test_check_input.py::TestAddInternalNodePrune::test_special_char_only_name
• tests/test_check_input.py::TestAddInternalNodePrune::test_internal_node_named_like_leaf
• tests/test_check_input.py::TestCheckSplice::test_single_gene_with_isoforms
• (no test yet for the sys.exit() exit status)

Proposed fix

One shared helper that lists input files (skip names starting with ., accept known extensions only) used by check-input and infer-roothogs, so both agree on what a proteome is; then fix the remaining cases individually. Bare except: around tree parsing should become specific exceptions with a non-zero exit.