fix(references): keep non-Latin reference labels

[lfazliev]

What

clean_label rejected any reference label without an ASCII alphanumeric
character ([A-Za-z0-9]), so Cyrillic and other non-Latin profile names —
common in RU/BY and other locales' people-search results — were silently
dropped. Relax the guard to the Unicode word-character check [^\W_]: a label
is kept when it has at least one letter or digit in any script, while pure
punctuation/symbol labels are still rejected.

Tests

• test_keeps_cyrillic_only_names — a Cyrillic-only name survives cleaning.
• test_rejects_punctuation_only_labels_across_scripts — punctuation-only
  labels are still dropped.

Synthetic prompt

In link_metadata.clean_label, the [A-Za-z0-9] guard drops valid non-Latin
(e.g. Cyrillic) profile names. Relax it to a Unicode word-character check so
any-script letters/digits pass while punctuation-only labels stay rejected.
Add regression tests.

[greptile-apps]

Greptile Summary

This PR updates reference label cleaning so non-Latin names are preserved. The main changes are:

• Adds a Unicode-aware label content check.
• Excludes invisible Hangul filler-only labels from valid reference text.
• Adds tests for Cyrillic names, punctuation-only labels, and invisible filler fallback.

Confidence Score: 5/5

This looks safe to merge.

• No blocking issues found in the changed code.

Important Files Changed

Filename	Overview
linkedin_mcp_server/scraping/link_metadata.py	Updates clean_label to keep Unicode letters and digits while rejecting filler-only invisible labels.
tests/test_link_metadata.py	Adds tests for non-Latin reference labels and invalid label fallback behavior.

_{Reviews (2): Last reviewed commit: "fix(references): reject invisible-only l..." | Re-trigger Greptile}

[stickerdaniel]

Hey, thanks for this! Rebased onto main and pushed one commit on top: [^\W_] also lets the four Hangul fillers through, so an invisible-only label passed and shadowed the aria_label fallback. Excluded them.