Paper: include token counts / corpus stats

[jowagner]

The reported sentence counts are not very useful to compare to other corpora as

• sentence splitting may be low quality

Suggestion: Also report token counts for current data

Related:

• issue #73

[jbrry]

Table 1 has been updated to show token counts for each corpus and the overall (171.3M).

• [x] Report token counts of current data

[ ] Report token counts of de-duplicated data --> now issue #105