Add invalid-input rejection tests for every encoding

[acul71]

Summary

• Fixes No test verifying invalid input is rejected for every encoding #41
• Adds tests/test_roundtrip.py that parametrizes over ENCODINGS and invalid suffixes (low-unicode, emoji, punctuation, high-latin1)
• Skips identity and base256emoji (as in the issue); expects DecodingError from the public decode() API

Demo / evidence

$ pytest tests/test_roundtrip.py -q
88 passed, 8 skipped in 0.11s

$ python -c "from multibase import decode, DecodingError; ..."
decode('f!'): DecodingError: Failed to decode multibase data: substring not found

Explanation: 22 encodings × 4 invalid suffixes = 88 assertions; 2 encodings × 4 = 8 skips. The smoke snippet shows alphabet garbage after a valid prefix raises DecodingError instead of succeeding.

Test plan

• pytest tests/test_roundtrip.py -q
• make lint
• make typecheck
• make test
• make docs-ci

Made with Cursor

[acul71]

Demo logs (invalid-input tests)

$ pytest tests/test_roundtrip.py -q
88 passed, 8 skipped in 0.11s

$ python - <<'PY'
from multibase import decode, DecodingError
try:
    decode("f!")
except DecodingError as e:
    print(f"{type(e).__name__}: {e}")
PY
DecodingError: Failed to decode multibase data: substring not found

88 cases = 22 encodings × 4 invalid suffixes; 8 skips = identity + base256emoji.

[acul71]

@sumanjeet0012 Could you please review this PR when you have a chance? Thanks!