Issue 67: Fix incorrect determination of APP1 size

This is a fix for issue #67 (closed) where adding new attributes can make the JPG unreadable as subsequent sections to the modified one get destroyed.

Most test baselines need to be updated as the padding bytes are now counted to be part of the APP1 section. Real changes appear only in add_short.py and little_endian_baselines/modify_ascii_same_len_hex.py where the start markers of the quantization table (FF DB) are now no longer part of the APP1 section, see #67 (closed) for details.

Unit test case that detects the issue is contained as well.

Also fixed some unrelated minor problems reported by pylint and sphinx.