Fix spectrum, tree, IO, and plotting bugs; add calculate_mutrate

[kpotoh]

Fixes several correctness bugs found while exploring the library (issue #13) and adds calculate_mutrate. Version bump: 0.0.15 → 0.0.16. All 97 tests pass.

Spectrum calculation (calculate_mutspec / MutRate)

• Keep RawMutSpec (obs/exp) separate from scaled MutSpec so scaling no longer overwrites the unscaled rates.
• All-zero scale no longer yields NaNs.
• Accept 12-component Mut values (C>T) as well as 192 (A[C>T]G).
• New calculate_mutrate() for unscaled observed/expected rates (exported from the package).
• collapse_mutspec accepts ObsNum/ExpNum as aliases of ObsFr/ExpFr and copies input to avoid SettingWithCopy.
• complete_sbs192_columns fills missing columns in one concat (no fragmentation warning).
• jackknife_spectra_sampling / calc_edgewise_spectra copy inputs instead of mutating caller frames.
• nobs_cutoff added; misspelled nobs_cuttof kept as an alias.
• filter_outlier_branches falls back to ProbaFull if ProbaMut is missing.

Codon annotation

• Sequences whose length is not divisible by 3: incomplete last codon is skipped; warn via warnings.warn (CodonAnnotation had no logger — that was a crash).
• _prepare_codontable now actually raises ValueError for invalid genetic-code arguments.
• get_syn_codons default is set() not dict().
• Empty mutation DataFrames get the expected columns.
• MutSpecExtractor._derive_mutspec now calls branch.process_branch() (serial extraction was broken after the parallel refactor).
• dump_table skips empty frames.

Tree

• TreeNode.parent / .up and iter_edges().
• Tree loads from a Newick string or a file path.
• node_parent catches StopIteration rather than BaseException.
• get_tree_height handles zero distances for geometric mean.
• calc_phylocoefs handles tree_height <= 0.
• get_ingroup_root no longer raises UnboundLocalError when both root children are leaves.

IO

• GenomeStates reads the states file (was using a leftover gappy-sites path); path_to_gappy_sites=None works.
• GenesStates passes states_fmt into read_alignment; parameterized SQL; groupby.size(); valid RuntimeError (illegal file= keyword removed).
• read_genbank_ref uses gene_qualifier and ftr.location.strand (BioPython 1.88 has no SeqFeature.strand).

Plotting and scripts

• plot_mutspec(show=False) with caller ax= no longer plt.close()s that figure (README multi-panel example was broken).
• scripts/calculate_mutspec.py: filter_short_exp_seqs actually drops IQR outliers; pivot(index=..., columns=..., values=...) for pandas 3.
• scripts/collect_mutations.py: empty dump handling; separate header flags for ms12 vs ms192 files.
• scripts/collect_mutations_parallel.py: same Branch.process_branch fix as the library.
• basic_logger() uses a named logger and does not attach duplicate handlers.

Tests

• tests/test_mutspec_calc.py now asserts real RawMutSpec values (old lookups never matched).
• New tests/test_bugfixes.py regression coverage.
• tests/test_codon_ann.py: real extract_mutations_simple test.

Suggested follow-ups (not in this PR)

These remain from issue #13 / README TODOs:

1. HGT annotation — new way of annotating mutations from HGT (no HGT source in this repo).
2. HGT parallelization — integrate parallel mutation extraction from the HGT project.
3. Split scripts — separate mutation collection vs annotation into distinct CLIs.
4. Merge tree.py into phylo_tree.py — existing TODO; node_parent / get_tree_height still live in a leftover module.
5. Seaborn boxplot vert deprecation — 204 MatplotlibDeprecationWarnings from seaborn internals vs matplotlib 3.11 (orientation instead of vert). Needs a seaborn upgrade or a local workaround in plot_mutspec.
6. Broader renames — nobs_cutoff is the start; remaining names like nobs_cuttof alias, tree_heigth typos in git history, and script --rewrite_db TODOs could be cleaned up in a dedicated pass.
7. Logging config — basic_logger still optionally reads a config file; inline the small config as noted in utils/logging.py.