docs: fix context_matching_mode param name and country-dir exceptions

Address Copilot review feedback:
- LemmaContextAwareEnhancer's constructor parameter is context_matching_mode,
  not matching_mode.
- The country_specific directory already contains a thai/ directory (and korea/),
  so 'only us and uk are exceptions' contradicted the layout. Reworded to describe
  the convention for new directories without blessing the pre-existing short forms.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013ck825cANTDYre5UCfmpYE

Author: claude

.github/copilot-instructions.md

@@ -60,9 +60,10 @@ class MyRemoteRecognizer(RemoteRecognizer):
 - NLP/ML-based: `.../predefined_recognizers/nlp_engine_recognizers/` or `.../ner/`
 - Third-party: `.../predefined_recognizers/third_party/`
 
-Directory names are the full lowercase country name (`south_africa`, `philippines`,
-`canada`), not the ISO country code. The only exceptions are the pre-existing `us`
-and `uk` directories. Do not add new abbreviated directories.
+Directory names should be the full lowercase country name (`south_africa`,
+`philippines`, `canada`), not the ISO country code. Some pre-existing directories
+predate this convention and use an abbreviation or short form (`us`, `uk`, `thai`);
+do not follow them. Do not add new abbreviated directories.
 
 Language codes are different: `supported_language` and the YAML `supported_languages`
 key take ISO 639-1 language codes (`ko` for Korean), not country codes (`kr`). A
@@ -105,7 +106,7 @@ cannot fire without context: `presidio-structured` counts matches per column and
 no surrounding context to work with.
 
 **Context words are matched as substrings.** `LemmaContextAwareEnhancer` defaults to
-`matching_mode="substring"`, so short context words fire on unrelated tokens.
+`context_matching_mode="substring"`, so short context words fire on unrelated tokens.
 
 ```python
 # ❌ BAD: "member" matches "remember", "auth" matches "author" and "OAuth",