Improve Opensearch standard tokenizer accent handling

[Dtenwolde]

Part of: https://github.com/duckdblabs/duckdb-internal/issues/9855

Couple of points in this PR:

• Reject strip_accents when the tokenizer mode is set to opensearch_standard
• Decomposed combining marks are now preserved inside tokens (these would previously cause a split)

A follow-up should work on:

• Improved word-boundary implementation using the ICU BreakIterator, but at the moment this is not compiled in the ICU extension.
• Opensearch-style ASCII folding (e.g. Straße --> strasse)
• Support for max_token_length

If we want even more compatibility, we should also look at the stemmer and stopword list to make it match with OpenSearch/Lucene