[alan-divd]

Summary

The example walks directories through self.iter(), which resolves each path
before recursing. Resolving follows symlinks, so a symlink pointing at one of
its own parents makes the walk descend the same tree over and over instead of
terminating.

Recurse through self._iter() instead, which skips the resolve step. The
entries come from iterdir() and are already known to exist, so resolving
them again was redundant regardless.

Also drops the path.joinpath(i), since iterdir() already yields full
paths.

Fixes #234

Test plan

• tox run -e py3 passes (518 passed)
• Walked a directory containing a symlink to its own parent. Before the
  change the walk repeats the same two paths indefinitely and never
  reaches the regular file next to the symlink; after it, the walk
  terminates and records the symlink as a link with its target