Function enrich_fs_error 

pub fn enrich_fs_error(path: &Path, op: &'static str, err: Error) -> Error

Wrap an io::Error produced by a filesystem operation against path with a heddle-context message naming both the operation and the path.

The mapping covers the cases users actually hit and the messages we promise from heddle’s CLI surface:

• ENOTEMPTY — usually remove_dir against a directory that still holds untracked or explicitly ignored content, such as build output. The high-level fix is to leave the directory in place, but when the error does surface (e.g. a path the planner did expect to remove), the message names the path so the user can investigate.
• EACCES — naming the path and the action (“removing”, “writing”, “renaming”) is enough for the user to inspect mode bits.
• ENOENT — caller-driven: only enriched when the operation expected the path to exist (so optional reads like a missing index pass through unchanged via the is_not_found predicate).
• EROFS — points the user at the filesystem mount, not at heddle.
• EXDEV — points the user at the temp path / mount mismatch.
• ENOSPC — same actionable disk-full message the snapshot path already relies on.

op is a verb in the present-progressive (“writing”, “removing”, “renaming”, “creating”) so the resulting message reads naturally: "could not remove because it contains content...".

The wrapped error preserves raw_os_error() (callers still classify disk-full via is_out_of_space) and exposes the original io::Error through the Error::source chain (so RUST_BACKTRACE=1 and anyhow’s chain printer still surface the OS error).