Added Hindi percentage ITN class

[mayuris-00]

Summary

Added a new percentage semiotic class to the Hindi ITN pipeline.
The system now converts spoken Hindi percentages to written form:

• बीस प्रतिशत → २०%
• सत्तर परसेंट → ७०%
• पाँच सौ फ़ीसदी → ५००%

Files Added

• hi/data/percentage/percent_symbol.tsv
• hi/taggers/percentage.py
• hi/verbalizers/percentage.py
• tests/nemo_text_processing/hi/test_percentage.py
• tests/.../test_cases_percentage.txt

Files Modified

• hi/taggers/tokenize_and_classify.py
• hi/verbalizers/verbalize.py

Test Results

All 12 percentage test cases passed.
All existing Hindi ITN tests still pass.

Verbose Trace

Input: बीस प्रतिशत
Tagged: tokens { percentage { integer: "२०" percent: "%" } }
Output: २०%

Input: सत्तर परसेंट
Tagged: tokens { percentage { integer: "७०" percent: "%" } }
Output: ७०%

[RajanPutty]

This file should not be here. You already have the correct copy at tests/nemo_text_processing/hi/test_percentage.py. This root-level version will not work -- the relative import from ..utils import CACHE_DIR requires the file to be inside the tests/nemo_text_processing/hi/ package. Please delete this file.

[RajanPutty]

This cdifflib fallback fix is unrelated to the percentage class. Even if it was needed to get your environment working, it should be a separate commit or a separate PR. Mixing unrelated fixes into a feature PR makes the review harder and the git history messier. Please remove this change from this PR and raise it separately if needed.

[mayuris-00]

Reverted the cdifflib change from this PR. Will raise it separately if needed.