keep break indicators [:-] during normalisation

All punctuation will be converted to '-'. Soft breaks : may be
added by preprocessors. The break signs are only used during
query analysis and are ignored during import token analysis.

src/nominatim_db/tokenizer/icu_token_analysis.py

@@ -25,6 +25,8 @@ class ICUTokenAnalysis:
 
     def __init__(self, norm_rules: str, trans_rules: str,
                  analysis_rules: Mapping[Optional[str], 'TokenAnalyzerRule']):
+        # additional break signs are not relevant during name analysis
+        norm_rules += ";[[:Space:][-:]]+ > ' ';"
         self.normalizer = Transliterator.createFromRules("icu_normalization",
                                                          norm_rules)
         trans_rules += ";[:Space:]+ > ' '"