Replacement of semicolon by visarga for normalization in Telugu creates invalid/misformed words

[ramSeraph]

The rule to replace ':' with 'ః" during normalization unconditionally, creates unicode sequences which are not valid.

Example: "హైదరాబాద్:" becomes "హైదరాబాద్ః", which is not a valid/wellformed Telugu unicode sequence.

This is the offending rule in the TeluguNormalizer:

text=re.sub(r'([\u0c00-\u0c7f]):','\\1\u0c03',text)