Remove fragments from Yoruba stop words (#14036)

[adedejimakinde]

Description

Fixes #14036.

The Yorùbá stop word list contained 24 entries that are not standalone words: three bare combining marks (U+0300, U+0301, U+0323), fourteen isolated consonants (b, d, f, g, j, k, l, p, r, s, t, w, y, ṣ), and seven vowels that only occur inside words (à, e, è, i, ì, u, ù). These look like the result of splitting text that wasn't Unicode normalised.

This PR removes them, taking the list from 91 to 67 entries. Single-character entries that are real Yorùbá words are kept, for example ó (he/she/it), ń (progressive marker), ò (negation), and the object pronouns á, é, í, ú.

It also adds spacy/tests/lang/yo/test_stop_words.py, which checks that every stop word is NFC normalised, that no entry starts with a combining mark, that the removed fragments stay out, and that real single-character words stay in.

I ran the new tests against spaCy 3.8.16 before and after the change: 25 failed and 5 passed with the original list, and all 30 passed with the fixed list.

Types of change

Bug fix (language data)

Checklist

• I confirm that I have the right to submit this contribution under the project's MIT license.
• I ran the new Yorùbá tests against spaCy 3.8.16 (all pass); the full suite will run in CI.
• My changes don't require a change to the documentation, or if they do, I've added all required information.