This commit updates `_parse_tree_structmap` so `FSEntry` objects are
created also when the parser runs into direct `fptr` elements, e.g.:

    <div TYPE="directory" LABEL="objects">
        <fptr FILEID="001.csv-0c0576c3-e82e-4501-9aef-1f3d2d35aeda"/>
        <fptr FILEID="002.mpg-722de890-2a3c-4343-bb69-f2b29894d9a7"/>
        <fptr FILEID="003.mpg-423247e7-c27d-4b5a-a64a-b02833f2212d"/>
    </div>

In Py2, ``FSEntry.path`` uses binary. This commit updates the constructor so
it uses the `utf-8` encoder when we're encoding text type. The default encoder
is `ascii` which is problematic.

In Py3, ``FSEntry.path`` is using Unicode and that needs to be fixed.

This avoids the following issue:

    > return ' '.join([str(x) for x in texts])
                       ^^^^^^
    E UnicodeEncodeError: 'ascii' codec can't encode characters in
      position 42-44: ordinal not in range(128)

Only fixed in Py2. Py3 codepath will need more work overall.