fix: shorten_middle output exceeds target width by ellipsis length

[nankingjing]

Problem

The shorten_middle function in src/kimi_cli/utils/string.py does not account for the
3-character "..." ellipsis when computing the left and right slice widths.
This causes the output to always be longer than the requested width by up to
3 characters.

Reproducer

>>> from kimi_cli.utils.string import shorten_middle
>>> shorten_middle("abcdefghij", 1)
'...j'       # length 4, expected at most 1
>>> shorten_middle("abcdefghij", 5)
'ab...hij'   # length 8, expected at most 5
>>> shorten_middle("abcdefghij", 8)
'abcd...hij' # length 11, expected at most 8

Root cause

Line 35: text[: width // 2] + "..." + text[-width // 2 :]

• The 3-character "..." is not subtracted from width, so the output is always
  width + 3 characters long (for width >= 3).
• For small widths where width < 3, Python's floor division of negative numbers
  (-width // 2) gives unexpected slice indices (e.g. -1 // 2 == -1, -2 // 2 == -1),
  producing even more misleading results.

Fix

Subtract 3 from the available width before computing half-widths, and handle the
edge case where width < 3 by falling back to a plain prefix slice:

available = max(0, width - 3)
left = available // 2
right = available - left
return text[:left] + "..." + text[-right:] if right > 0 else text[:width]

Call site impact

The known caller (src/kimi_cli/tools/__init__.py:97) uses width=50, where the
current code would produce 53-character output instead of 50.

Verification

Tested locally with the same Python file:

>>> def fixed_shorten_middle(text, width, remove_newline=True):
...     if len(text) <= width:
...         return text
...     import re
...     if remove_newline:
...         text = re.sub(r"[\r\n]+", " ", text)
...     available = max(0, width - 3)
...     left = available // 2
...     right = available - left
...     return text[:left] + "..." + text[-right:] if right > 0 else text[:width]
...
>>> fixed_shorten_middle("abcdefghij", 5)
'a...j'       # length 5, within width ✅
>>> fixed_shorten_middle("abcdefghij", 1)
'a'           # length 1, within width ✅
>>> fixed_shorten_middle("abcdefghij", 8)
'ab...hij'    # length 8, within width ✅
>>> fixed_shorten_middle("abcdefghij", 50)
'abcdefghij'  # already short enough, returned as-is ✅

---

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 1 additional finding.