fix(shell): adapt timeouts for long commands - #2200

Rebased this onto the latest main and resolved the shell-tool conflict by keeping both recent behaviors: Windows command preprocessing still runs before execution, and long-running commands now compute the adaptive timeout from the same command string that is approved and executed.

Validation on head bbadafd9:

• python -m uv run pytest tests\tools\test_shell_timeout_policy.py -q (5 passed)
• python -m uv run pytest tests\tools\test_shell_bash.py tests\tools\test_shell_timeout_policy.py -q (5 passed, 24 skipped)
• python -m uv run ruff check src\kimi_cli\tools\shell\__init__.py tests\tools\test_shell_timeout_policy.py
• python -m py_compile src\kimi_cli\tools\shell\__init__.py tests\tools\test_shell_timeout_policy.py
• git diff --check upstream/main..HEAD

Pytest still reports a local .pytest_cache permission warning in this checkout, but the tests completed successfully.

Gentle bump on this one. The long-command timeout adaptation still applies cleanly on current main, and it is a small self-contained change. Happy to rebase if it is useful to pick up.

Third and final nudge. src/kimi_cli/tools/shell/__init__.py is untouched on current main, so long-running commands still hit the fixed timeout and the adaptive policy here still applies. Glad to adjust the approach if you would rather see this handled somewhere else.

The repo README announces the wind-down into kimi-code and this has had three nudges with no review. Closing rather than pinging a fourth time.