Some improvements on cognate set detection

[Anaphory]

Also the final fixes for multiple languages in one long format table.
Sorry that my feature branches are so messy.

[codecov]

Codecov Report

Merging #369 (4c565b6) into master (fced3f3) will increase coverage by 0.23%.
The diff coverage is 91.57%.

@@            Coverage Diff             @@
##           master     #369      +/-   ##
==========================================
+ Coverage   80.99%   81.23%   +0.23%     
==========================================
  Files          43       43              
  Lines        4326     4412      +86     
  Branches     1098     1129      +31     
==========================================
+ Hits         3504     3584      +80     
- Misses        629      632       +3     
- Partials      193      196       +3     

Impacted Files	Coverage Δ
src/lexedata/importer/excel_long_format.py	89.44% <83.33%> (-0.60%)	⬇️
src/lexedata/edit/detect_cognates.py	59.78% <92.77%> (+26.44%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fced3f3...4c565b6. Read the comment docs.