Describe the bug

GreedyEncoder in pyaptamer/trafos/encode/_greedy.py has two failure modes
when initialized with an empty words dict, depending on whether word_max_len
is provided:

Mode 1 — Confusing crash when word_max_len=None (the default):

import pandas as pd
from pyaptamer.trafos.encode import GreedyEncoder

enc = GreedyEncoder(words={})
X = pd.DataFrame({"seq": [["A", "C", "G"]]})
enc.fit_transform(X)
# ValueError: max() iterable argument is empty

The user has no idea the root cause is words={}. The error
points to an internal max() call, not the bad input.

Mode 2 — Silent all-zeros output when word_max_len is provided:

enc = GreedyEncoder(words={}, word_max_len=3)
X = pd.DataFrame({"seq": [["A", "C", "G"]]})
result = enc.fit_transform(X)
print(result)
#    0  1  2
# 0  0  0  0
# No error, no warning — silently returns garbage output

Expected behavior

An empty words dict should raise a clear ValueError at initialization,
consistent with how the rest of the codebase handles invalid inputs.

GreedyEncoder(words={})
# ValueError: `words` must not be empty.

Additional context

• Affected file: pyaptamer/trafos/encode/_greedy.py
• Fix: add an empty dict guard in __init__ with a Raises section in the docstring
• A test should be added to the existing encoder test suite