Simplify pass over the changes since v1.2.1

[derek73]

What changed

A /simplify review (four parallel angles: reuse, simplification, efficiency, altitude) over everything since the v1.2.1 tag, with the surviving findings applied. All changes are behavior-preserving — net −34 lines.

nameparser/parser.py

• join_on_conjunctions: both join branches carried identical 5-line title/prefix registration blocks (including the pre-existing title duplication); they now share one local register_joined_piece helper.
• The lenient "post-comma suffix_not_acronyms members bypass the is_an_initial veto" rule was encoded twice; it now lives in one new predicate, is_suffix_lenient(). are_suffixes_after_comma is a one-line all() over it, and the single-use 20-line is_suffix_at_lastname_comma_end method is deleted, with its positional guards (nxt is None and len(parts) == 2) inlined at the lastname-comma loop that owns that context.
• cap_word: computes its exception-lookup key once instead of up to four lc()/.replace() string allocations per word (runs per word-match on every capitalize() pass).

nameparser/config/__init__.py

• Constants.__setstate__: dropped the 10-line verification loop — its condition can never fire for any state __getstate__ produces, and removing it also eliminated the duplicated descriptor-scan between the two pickle hooks.
• The dunder guard copy-pasted across four TupleManager/RegexTupleManager attribute hooks is now one _is_dunder() helper carrying the shared rationale; site-specific failure notes stay at each hook.

Tests

• New FlaggedConstantsTestBase (with a constants_kwargs class attribute) replaces the identical setup_method/hn() pair copy-pasted into 7 classes across the patronymic-order and middle-name-as-last test files.
• Deleted the two test_prefixes.py invariant tests that duplicate the import-time asserts in prefixes.py — a violated assert prevents import entirely, so the tests could never be the thing that fails.

AGENTS.md's note on the suffix/initial tension is updated to describe is_suffix_lenient().

Reviewer notes

• Skipped as behavior-changing (not applied): the bound-first-name join is missing from the suffix-comma branch of parse_full_name — "Abdul Salam Hassan, MD" parses differently from "Abdul Salam Hassan MD". Tracked separately rather than folded into this cleanup PR.
• Also skipped: moving the suffix_delimiter split to suffix-classification time (design change), and extracting parse_nicknames' intentionally-inlined suffix normalization (the shared version came out clunkier than the commented inline form).

Verification

• uv run pytest: 1236 passed, 22 xfailed (same xfail count as before)
• uv run mypy nameparser/: clean; tests-included mypy error count identical to the pre-change baseline (41)

🤖 Generated with Claude Code