Raise NumOutOfRangeError from tenfn instead of leaking IndexError (#242)

[chiliec]

Fixes #242.

Problem

number_to_words documents (and imports/advertises) NumOutOfRangeError for out-of-range inputs, but for numbers requiring a magnitude index beyond the largest scale word it leaks an uncaught IndexError instead:

import inflect
p = inflect.engine()
p.number_to_words(10**40)
# Expected: NumOutOfRangeError
# Actual:   IndexError: list index out of range  (from tenfn)

10**36 raises NumOutOfRangeError correctly; 10**40 and above do not.

Root cause

millfn performs the bounds check and raises NumOutOfRangeError:

def millfn(self, ind: int = 0) -> str:
    if ind > len(mill) - 1:
        raise NumOutOfRangeError
    return mill[ind]

Every scale-word lookup goes through it — except the teen branch of tenfn (tens == 1), which indexed the table directly:

return f"{teen[units]}{mill[mindex]}"   # unchecked

so an out-of-range mindex raised IndexError from that line rather than the documented error.

Fix

Route the lookup through millfn, matching every other path:

return f"{teen[units]}{self.millfn(mindex)}"

Tests

• test_tenfn: the teen branch now raises NumOutOfRangeError for an out-of-range magnitude.
• test_issue_242: number_to_words(10**36 / 10**40 / 10**100) all raise NumOutOfRangeError, and number_to_words(10**33) still returns "one decillion".

Adds a newsfragments/242.bugfix.rst.

Verification (local, Python 3.11)

• RED→GREEN: with the fix reverted, both new tests fail with IndexError at the tenfn line; with the fix they pass.
• Full suite: pytest — 215 passed, 16 xfailed, no regressions.
• ruff introduces no new findings on the changed line.

[CAOShurong]

Verified independently (no run, no claim). I cloned the repo and checked out the PR head 3db76f5d (3db76f5dbfe1526fbc0cb05b251b3abc3b9b5d77).

Root cause matches the PR description exactly: the teen branch of tenfn indexed mill[mindex] directly instead of routing through millfn() (which is the only path that raises the documented NumOutOfRangeError). So inputs whose magnitude index exceeds len(mill) - 1 leaked an uncaught IndexError instead of the advertised error type.

Reproduction (RED on base 262a247d):

• inflect.engine().number_to_words(10**40) raises IndexError: list index out of range at inflect/__init__.py:3699 (the mill[mindex] line).
• New tests/test_numwords.py::test_issue_242 FAILS on base with that same IndexError.

GREEN at the PR head:

• git checkout 3db76f5d → tests/test_numwords.py::test_issue_242 and the tests/test_pwd.py tenfn assertion both pass.
• Full suite via uv run --with pytest --with more-itertools python -m pytest tests/: 172 passed, 16 xfailed, no regressions (the new test_issue_242 + test_pwd additions included).

The fix is a 1-line change that routes the lookup through self.millfn(mindex), consistent with every other scale-word path, and it adds a newsfragments/242.bugfix.rst changelog entry. Small, correct, and well-tested. Approved.

(AI assistance used to collect verification evidence; the fix and tests are the author's.)