fix: improve numeric literal validation

[smortezah]

This pull request improves numeric literal detection and ensures consistency between encoder and decoder utilities. It refines the is_numeric_literal function to properly handle negative numbers with leading zeros, and introduces a comprehensive test suite to verify numeric detection logic and round-trip encoding/decoding behavior.

Numeric literal detection improvements

• Updated is_numeric_literal in src/toon_format/_literal_utils.py to correctly reject negative numbers with leading zeros (e.g., -01), ensuring stricter validation of numeric literals.

Test coverage enhancements

• Added tests/test_numeric_detection.py with extensive tests for both is_numeric_literal and is_numeric_like, covering valid numbers, floats, scientific notation, leading zeros, non-numeric strings, and special float values.
• Included consistency tests to verify that encoder and decoder utilities agree on valid and invalid numeric representations, especially for octal-like numbers with leading zeros.
• Added round-trip tests to confirm that encoding and decoding preserve the correct types for valid numbers and octal-like strings, preventing unintended type conversions.

[bpradana]

approved!
@Justar96 is it okay if I implement git submodules for the fixtures? So the fixtures are always up to date.
What do you think? 😄

[Justar96]

@bpradana pls do. Thanks a lot

[davidpirogov]

Minor update to test asserts, otherwise looks good to me