organize post-processing in a maintainable way

[S6savahd]

for each volume, we will have 0-3 things that the fully automated case gets wrong.
For these, we need hard-coded post-processing rules and these rules should be in a place where it is easy to maintain them.
For example, we do not want 100s of lines of volume specific text distinctions in serializer.py.
(maybe with a subdirectory of files with Python modules).

[S6savahd]

it is good in its current status but since this is a longterm issue, we keep it open.

[liyakun]

changed all previous text processing by using triple.