Add comprehensive tests: coverage from 54% to 97%

[Copilot]

Summary

This PR significantly improves the test coverage of pyCrossTalkeR from 54% to 97% by adding 75 new tests and fixing two pre-existing bugs found during test writing.

New files

File	Description
test/helpers.py	Shared SEL_COLUMNS constant and make_lr_df() synthetic-data factory
test/__init__.py	Makes test/ a proper package so helpers.py is importable
test/conftest.py	Session-scoped analysed_adata pytest fixture (synthetic CTR + EXP conditions)
test/test_coverage.py	75 new tests across 20+ test classes

Coverage per module

Module	Before	After
tools/utils.py	71%	99%
tools/Single_Condition.py	81%	100%
tools/Comparative_condition.py	77%	100%
plots/plot.py	21%	93%
TOTAL	54%	97%

What is tested

• utils.py: add_node_type, graph JSON serialization (including numpy scalar edge attrs), get_clustered_node_order, create_ordered_circular_layout, ranking_net (signed + unsigned modes), comparative_pagerank/comparative_med, fisher_test_cci and mannwhitneyu_test_cci with explicit comparison parameter, from_liana (DataFrame path with label, dict path, pval_filter, compute_means)
• Single_Condition.py: dict-of-CSV-paths input, invalid input types (ValueError), receptor-first gene ordering
• Comparative_condition.py: explicit comparison parameter, multiple comparison pairs
• plots/plot.py: plot_cci with log=True, single-coord, emax; plot_bar_rankings with all type filter lengths (1/2/3 chars), filter_sign pos/neg, empty result message; plot_clustermap and plot_graph_clustermap; plot_volcane with annotated red points; plot_pca_LR_comparative (non-ggi, ggi with gene_types, include_tf=True); plot_sankey all filter branches; gen_sankey2

Bug fixes

1. plot_bar_rankings filter_sign (plots/plot.py line 337–339): Referenced hardcoded string 'ranking' instead of the ranking parameter, always raising KeyError. Fixed to use the variable.

2. plot_pca_LR_comparative include_tf=True (plots/plot.py line 195): Used pd.DataFrame(pca_df.index, columns=["gene"]) which returns an empty DataFrame in pandas 2.x when the index originates from a .loc[] call. Fixed to pd.DataFrame({"gene": pca_df.index}).

Remaining uncovered lines (3%)

• utils.py 458, 509: Dead code — B <= 0 branch, B is hardcoded to 1000
• plot.py 187: gene_types="TF" in ggi mode produces empty PCA → crash
• plot.py 227: PCA text annotation only fires when a point is >4σ (not reproducible with small synthetic data)
• plot.py 415, 421: Legacy data handling when gene names don't yet have | separators
• plot.py 683–716: gene_annotation() — requires live external gProfiler API call
• generate_report.py 80: save=True branch — requires a filesystem output path