Add jobs to study alignments

[Icemole]

PlotViterbiAlignmentJob: job to plot the Viterbi alignment. The point of this job is to check for a specific issue in the final alignments.

---

DumpSegmentTextAlignmentJob: job to dump text/alignment pairs for a specific corpus and set of alignment files. The point of this job is to review in a fast way the text against the corresponding alignments.

The output is a compressed text file of the form:

<segment-id>
<segment-text>
<id-0> <timestamp-start-0> <timestamp-end-0> <alignment-0> <weight-0>
...

Feel free to propose any other format in the comments.

[DanEnergetics]

I prefer the second to last or last dumping format that you suggested. They would make it easiest to get a quick grasp on the duration of certain phonemes for me.

[Icemole]

I prefer the second to last or last dumping format that you suggested. They would make it easiest to get a quick grasp on the duration of certain phonemes for me.

Maybe we can specify a default mapping (csv file?) and an optional custom mapping to be provided by the user.

Edit/update: I tried to do this but faced issues with detailed print schemas: if we want to print timestamps and so on, we should provide the alignments as a tuple (timestamp, allophone_id, hmm_state, weight), the string corresponding to simply the triphone + HMM state doesn't do it. Of course, this could be easily solved by adding the specific triphone to the tuple, but that tuple format wouldn't be consistent with the rest, and I'd like the user to know what they're doing.

Let me know if you still want this custom printing approach.