Give _FracTrim a __repr__ so format errors raise ValueError

[gaoflow]

Problem

In the pure-Python implementation, a format-pattern error that references a
trimmed-fraction (F) field crashes with AttributeError instead of the
intended ValueError:

from whenever import Time, Date

Time(1, 2, 3, nanosecond=4).format("fF")   # duplicate nanos field
# AttributeError: '_FracTrim' object has no attribute 'pattern'
Date(2024, 3, 15).format("FFF")            # F unsupported for a date
# AttributeError: '_FracTrim' object has no attribute 'pattern'

_FracTrim inherits _Field.__repr__, which reads self.pattern. _FracTrim
never sets that instance attribute (pattern is a class-only annotation), so
rendering such a field inside an error message — duplicate-field detection or
unsupported-field rejection — raises AttributeError rather than the documented
ValueError. The Rust extension already renders these correctly, so this is a
pure-Python/Rust divergence.

Fix

Give _FracTrim a __repr__ returning its letter repeated by its width (e.g.
"FFF"), mirroring the sibling _FracExact. The error paths then render and
raise ValueError as intended.

Tests

TestFracTrimErrorRendering covers the __repr__ output and asserts a
ValueError (not AttributeError) for duplicate-nanos and unsupported-F
cases across compile_pattern, format, and parse.

Pure-Python suite tests/test_format_parse.py → 225 passed. ruff check clean.

[ariebovenberg]

@gaoflow thanks for taking the time to post this issue and fix it.

It makes me wonder if there are other patterns where this pops up 🤔

[ariebovenberg]

Looks like only F format code was affected. Thanks again for the fix.