fix: guard os.getcwd() against deleted CWD (FileNotFoundError)

[isair]

When a scratch workspace is cleaned up while a terminal session is active, os.getcwd() raises FileNotFoundError and the tool crashes. Wrap it in try/except and fall back to TERMINAL_CWD env var or the user's home directory.

This is a common edge case in CI, ephemeral containers, and workspace-tooling environments where directories may be deleted out from under a running process.

[alt-glitch]

Duplicate of #33377 — same fix at the same call site (os.getcwd() guard in _get_env_config() in tools/terminal_tool.py, FileNotFoundError fallback when CWD is deleted). #33377 is the open canonical PR; #33407 was a closed earlier copy. Consolidating there.

[Morad37]

One more spot to cover: the Docker mount path still calls os.getcwd() in the docker_mount_cwd_to_workspace branch when TERMINAL_CWD is unset.

So this fixes env_type=local, but env_type=docker with TERMINAL_DOCKER_MOUNT_CWD_TO_WORKSPACE=true can still hit the same FileNotFoundError if the process CWD was deleted. Reusing the guarded helper there would close the same edge case for mounted workspaces too.

[teknium1]

Merged via #39491 (commit ad69d3e) with your authorship preserved via rebase. Your deleted-CWD guard landed, refactored into a _safe_getcwd() helper applied to all three os.getcwd() sites in the file (local default, Docker cwd-passthrough, debug print) so the same crash cant resurface at a sibling. Thanks!