Config file with UTF-8 BOM causes Invalid TOML error on startup

Bug Description

When config.toml contains a UTF-8 BOM (Byte Order Mark) at the beginning, kimi fails to start with:

Invalid TOML in configuration file ~/.kimi/config.toml: Empty key at line 1 col 0

Steps to Reproduce

1. Edit ~/.kimi/config.toml with a text editor that adds BOM on save (e.g. Windows Notepad, or certain PowerShell Out-File operations).
2. Run any kimi command.
3. The CLI crashes immediately with the above TOML parse error.

Root Cause

src/kimi_cli/config.py uses tomlkit.loads() to parse the config file, but tomlkit does not handle the UTF-8 BOM (\xef\xbb\xbf) that some Windows editors prepend to UTF-8 files. The BOM bytes are interpreted as part of the first key name, leading to the "Empty key" error.

https://github.com/MoonshotAI/kimi-cli/blob/main/src/kimi_cli/config.py#L298

config_text = config_file.read_text(encoding="utf-8")
data = tomlkit.loads(config_text)  # fails if config_text starts with \ufeff

Proposed Fix

Strip the BOM before parsing:

config_text = config_file.read_text(encoding="utf-8")
if config_text.startswith("\ufeff"):
    config_text = config_text[1:]
data = tomlkit.loads(config_text)

Environment

• OS: Windows 11
• Editor: Windows Notepad / VS Code with "UTF-8 with BOM" encoding
• kimi-cli version: latest

Start in src/kimi_cli/config.py around the configuration loading at line 298, and inspect how config.toml is read before tomlkit.loads() parses it. Reproduce with a UTF-8 BOM-prefixed file, then verify that kimi starts successfully and still reports genuinely invalid TOML errors.