Only tok2vec, tagger, and sentence splitting are used (pos_tag and
sent_tokenize). Exclude ner, parser, lemmatizer, attribute_ruler when
loading en_core_web_sm, and add lightweight sentencizer to replace the
dependency parser for sentence boundary detection.

Saves ~12 MiB of model weights per process.

Per review feedback, removing parser and using sentencizer causes
sentence splitting regressions. Keep parser loaded, only exclude
ner, lemmatizer, and attribute_ruler.