Fix dedupe() crash when an item is emptied by the processor

[Sanjays2402]

Fixes #94

The bug

process.dedupe() crashes with ValueError: max() iterable argument is empty on otherwise valid input when one of the items is reduced to an empty string by the default processor:

from thefuzz.process import dedupe
dedupe(['###', 'apple', 'apple pie'])
# ValueError: max() iterable argument is empty

(utils.full_process('###') yields ''.)

Root cause

dedupe() calls extractBests() per item and then max() over the returned matches:

for item in contains_dupes:
    matches = extractBests(item, contains_dupes, scorer=scorer, score_cutoff=threshold, limit=None)
    deduped.add(max(matches, key=lambda x: (len(x[0]), x[0]))[0])

When an item's processed form is empty, every comparison scores 0, so at the default threshold=70 extractBests() returns [] — not even the item itself matches. max([]) then raises. The code even emits a warning that a processor-emptied query makes all comparisons score 0, but still calls max() unconditionally.

Confirmed with the real code:

>>> extractBests('###', ['###', 'apple', 'apple pie'], score_cutoff=70, limit=None)
[]                       # -> max([]) raises
>>> extractBests('apple', ['###', 'apple', 'apple pie'], score_cutoff=70, limit=None)
[('apple', 100), ('apple pie', 100)]

The fix

An item with no matches has no fuzzy duplicates, so keep it as its own cluster (add the item itself to the deduped set) instead of crashing. This is exactly the behavior the issue asks for, and leaves the result for every non-empty item unchanged.

matches = extractBests(item, contains_dupes, scorer=scorer, score_cutoff=threshold, limit=None)
if not matches:
    deduped.add(item)
    continue
deduped.add(max(matches, key=lambda x: (len(x[0]), x[0]))[0])

Before/after:

call	before	after
dedupe(['###', 'apple', 'apple pie'])	ValueError	['###', 'apple pie']
dedupe(['###', '!!!', '@@@'])	ValueError	['###', '!!!', '@@@'] (each kept)
dedupe(['Frodo Baggin', 'Frodo Baggins', 'F. Baggins', ...])	works	unchanged
dedupe(['Tom', 'Dick', 'Harry']) (no dupes)	returns original list	unchanged

Regression test

Added test_dedupe_with_empty_processed_item, which reproduces the crash and asserts the empty-processed item is preserved while the genuine duplicates still collapse. Proof it guards the bug:

• Without the source fix (test present): FAILED ... ValueError: max() iterable argument is empty at thefuzz/process.py:441.
• With the fix: 1 passed.

Full suite green: 50 passed in test_thefuzz.py (including the TestCodeFormat pycodestyle gate) and test_thefuzz_pytest.py passes. The change adds no new pycodestyle errors in thefuzz/ (baseline and post-change both 0 non-E501). Diff is +30/−0 across thefuzz/process.py and test_thefuzz.py.