Implement backwards incompatible changes

[padix-key]

This PR implements the currently open 'major release' issues and removes all deprecated functionality from the code base.

Issue	Title	Implementation
#947	Remove license banner from source files	Removed the three-line BSD banner (and the following blank line) from all 346 .py, .pyi and .rst files.
#703	Make EValueEstimator accept a Generator	EValueEstimator.from_samples() got an rng: numpy.random.Generator parameter (default: fresh default_rng()); sampling uses rng.choice(). Tests and gallery scripts pass a seeded generator instead of np.random.seed().
#905	Do not introduce line breaks in FastaFile by default	chars_per_line of FastaFile, FastaFile.read() and FastaFile.write_iter() accepts None and defaults to it, analogous to FastqFile. Internal callers that used a huge line width now rely on the default.
#923	Make FastqFile offset an IntEnum	FastqFile.Offset already existed; the docstrings still advertised string offsets and the docs still passed offset="Sanger". Docstrings now document the enum members, and tutorial/gallery scripts use FastqFile.Offset.SANGER.
#900	Remove verbose parameter from fetch() functions	Removed the parameter, the progress print() calls and the docstring entries from rcsb, afdb, uniprot, entrez and pubchem fetch().
#899	fetch() functions should accept and return Path objects	All fetch() functions (and entrez.fetch_single_file()) accept str | PathLike targets, create directories via Path.mkdir() and return pathlib.Path objects (or lists thereof). Tutorials use .name instead of os.path.basename().
#930	Change default scoring scheme for alignment functions	align_ungapped(), align_optimal(), align_banded(), align_local_gapped(), align_multiple() and score() default to matrix=(1, -1) and gap_penalty=-1; score() accepts (match, mismatch) tuples. To make matrix optional, band (in align_banded()) and seed/threshold (in align_local_gapped()) now precede matrix. Doctest examples that illustrate specific alignments pin gap_penalty=-10. The Rust SeedExtension constructor and the deprecated align_local_ungapped() are unchanged.
#924	Represent secondary structure elements as IntEnum	New biotite.structure.PSeaElement (NONE=-1, COIL, HELIX, STRAND) returned by annotate_sse() and pdbx.get_sse(), and DsspElement (COIL … POLYPROLINE_HELIX) returned by the v1 and v2 DsspApp. Both enums provide symbol, from_symbol(), from_symbols() and to_symbols() for one-letter conversion.
#909	Represent Compiled Graphics Objects using dedicated class	New CGO class wrapping the float values (values property, __eq__, __len__); the former directive enum became CGO.Type. get_*_cgo() return CGO, draw_cgo() takes Sequence[CGO] and rejects raw lists.
#908	Add ColorScheme	Frozen dataclass ColorScheme(name, alphabet, colors) with fit(alphabet, default). load_color_scheme() returns it; LetterTypePlotter, plot_alignment_type_based() and plot_sequence_logo() accept it as color_scheme/scheme. get_color_scheme() keeps returning a color list.
#906	Add GFFRecord	Dataclass GFFRecord(seqid, source, type, start, end, score, strand, phase, attributes). GFFFile indexing/iteration yields records, insert()/append()/__setitem__() take a record; converters updated.
#925	Add GenBankLocus	Dataclass GenBankLocus(name, length, mol_type, is_circular, division, date). get_locus() returns it and set_locus(gb_file, locus) accepts it.
#907	Add GenBankRecord	Dataclass GenBankRecord(name, content, subfields). GenBankFile indexing/iteration and get_fields() return records; insert()/append()/set_field()/__setitem__() take a record. The metadata, annotation and sequence converters use record attributes.
#902	Use dataclasses to better encapsulate large tuples and untyped dictionaries	Umbrella issue, fulfilled by the ColorScheme, GFFRecord, GenBankLocus, GenBankRecord and CGO commits above; no separate commit.
#844	Refactor AffineTransformation	Renamed to RigidTransformation(rotation, translation) using a single rotation and translation (x' = R x + t); __mul__ composes transformations ((a * b).apply(x) == a.apply(b.apply(x))) and unary __neg__ inverts them. apply() also handles single Atoms. superimpose*(), space_group_transforms() and PDBx assembly code produce it; align_vectors() and orient_principal_components() now return the transformation instead of transformed coordinates.
–	Rename application_v2 to application	biotite.application_v2 was moved to biotite.application (sources, tests, docs, CI config). A deprecated biotite.application_v2 shim remains: importing it emits a DeprecationWarning, re-exports biotite.application and aliases all subpackages in sys.modules, so biotite.application_v2.dssp.DsspApp is the same object as biotite.application.dssp.DsspApp. The shim is excluded from the API documentation.
–	Remove deprecated code	Removed the deprecated v1 biotite.application package (and its tests, doctests and API docs; the former application_v2 was left untouched in this commit), align_local_ungapped(), set_font_size_in_coord(), PDBFile.get_symmetry_mates()/pdb.get_symmetry_mates(), the as_bytes parameter of LetterAlphabet.decode*(), the distance_range parameter of connect_via_distances(), the float_tolerance parameter of pdbx.compress(), the include_bonds parameter of pdbx.set_structure() and its label_entity_id annotation fallback, and the as_mask parameter of CellList.get_atoms*() (Rust + stub). plot_nucleotide_secondary_structure() and the MUSCLE reference tests now use the v2 apps.
#773	Make use_author_fields default to True (i.e. use label fields)	Default of use_author_fields in get_structure(), get_assembly() and get_unit_cell() flipped to False. get_sequence() keys its result by struct_asym.id (= label_asym_id), get_sse() uses beg_label_asym_id/label_asym_id. Tests and docs that compare against PDB/PDBQT/PyMOL/foldseek data (which only know author fields) or select chains by author IDs (2RTG, 2OR1) request use_author_fields=True explicitly; #553 was already fixed.

Closed issues

Closes #703.
Closes #905.
Closes #923.
Closes #900.
Closes #899.
Closes #930.
Closes #924.
Closes #909.
Closes #908.
Closes #906.
Closes #925.
Closes #907.
Closes #902.
Closes #844.
Closes #773.

[codspeed]

Merging this PR will regress 2 benchmarks

⚡ 9 improved benchmarks
❌ 2 regressed benchmarks
✅ 98 untouched benchmarks
⏩ 14 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	benchmark_structural_alphabet_methods[to_protein_blocks]	6.6 ms	10.2 ms	-35.14%
❌	benchmark_match_kmer_selection[KmerTable-None]	246.2 µs	277.8 µs	-11.38%
⚡	benchmark_set_structure[cif-False]	92.9 ms	31.6 ms	×2.9
⚡	benchmark_set_structure[cif-True]	103.5 ms	41.7 ms	×2.5
⚡	benchmark_connect[connect_via_distances]	3.2 ms	1.7 ms	+86.35%
⚡	benchmark_serialize_pdbx[cif]	177.7 ms	120.5 ms	+47.38%
⚡	benchmark_infer_bond_types	1,203 µs	830.7 µs	+44.82%
⚡	benchmark_clustering[neighbor_joining]	557.7 µs	394.5 µs	+41.35%
⚡	benchmark_set_structure	9.4 ms	7 ms	+35.39%
⚡	benchmark_set_structure_with_bonds	9.6 ms	7.2 ms	+34.53%
⚡	benchmark_match_kmer_selection[BucketKmerTable(10000)-None]	353.1 µs	320.9 µs	+10.04%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing padix-key:v2-finish (32f645d) with main (1aaf784)²}

Footnotes

1. 14 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on v2 (b314fa7) during the generation of this report, so main (1aaf784) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩