fix(nlp): match single-letter bullets in the enumerated bullet pattern #4635

ENUMERATED_BULLETS_RE cannot match a single-letter bullet.

unstructured/nlp/patterns.py:111 compiles

ENUMERATED_BULLETS_RE = re.compile(r"(?:(?:\d{1,3}|[a-z][A-Z])\.?){1,3}")

[a-z][A-Z] is two single-character classes placed next to each other, so it matches exactly two characters (aB) and never a single a or A. The two ordered-bullet helpers in this repo both accept a single letter as a marker:

• unstructured/cleaners/core.py:53 clean_ordered_bullets — docstring says "accounting numeric and alphanumeric types" and shows a.b This is a very important point -> This is a very important point.
• unstructured/cleaners/extract.py:104 extract_ordered_bullets — docstring shows a.1 This is a very important point -> ("a", "1", None).

The splitter regex used by the list path cannot, so lettered lists come out as one blob:

from unstructured.partition.common.common import layout_list_to_list_items
[e.text for e in layout_list_to_list_items("a. foo\nb. bar", None, None, None, None)]
# ['a. foo\nb. bar']   expected ['foo', 'bar']

This is user visible through partition_image and partition_pdf(strategy="ocr_only"): ocr_data_to_elements -> normalize_layout_element -> layout_list_to_list_items (unstructured/partition/common/common.py:137).

Change

Keep the numeric branch byte for byte, and add a letter branch that only matches at the start of a line, only with a literal dot, and only when whitespace or the end of the text follows. The line anchor and the (?!\s[a-zA-Z]\.) guard are both required: \.? is optional, so an unanchored [a-zA-Z] turns every letter into a separator ("1. foo\n2. bar" would split into []), and without the guard "e. g. spaced example" loses its first word.

Verification

test_unstructured/nlp/test_patterns.py (new) covers both the regex and the converter, including the existing numeric behaviour and the abbreviation cases:

python -m pytest test_unstructured/nlp/test_patterns.py test_unstructured/nlp/test_tokenize.py test_unstructured/cleaners/test_core.py test_unstructured/cleaners/test_extract.py -q

pristine: 19 failed, 211 passed
patched:  10 failed, 220 passed

The 10 remaining failures are the pre-existing test_tokenize.py ones (no spaCy model in this offline environment); they are identical in both runs. The new file alone goes from 9 failed, 11 passed to 20 passed.

ruff format --check and ruff check are clean on both files.