new command for automatic extraction of transcription data with manual override

[LinguList]

We need a new schema to make normal data from transcripton-data-sources. For this, I recommend:

• reformat the sources in such a way that we add a column for BIPA (this can serve as an explicit override)
• read in teh data from a command in __main__
• copy-paste unlinked data as well, as we'll assign it to a special group "NA", as we do in concepticon

In this way, people can contribute TD by just submitting a text-file with BIPA-counterparts (suggested), or without. If counterparts are given, we'll take them at face value, if not we try to interpret the data.

[LinguList]

Question remains: how to distinguish the three cases as indicated in #89, namely:

• explicit sound defined as a mapping in a dataset
  • bipa sound is existing
  • bipa sound is not existing (thus generated)
• automatic mapping
  • automatic mapping yields plain bipa sound
  • automatic mapping yields generated sound

Unknown sounds will always be excluded, due to a new check. Maybe best go for a boolean value for each of the two situations, so we have the value "generated" (referring to the BIPA symbol), and some column "explicit" (referring to the mapping).