asciifying http header for csv download; fixes #3952

[rumbin]

I think this should be all that is needed to fix #3952.

This way, the rather pathologic tab label →ʬéıρδ Ñämë← will get converted to WWeird_Name, which is pretty much what I would expect it to be in non-unicode representation.

[xrmx]

Also a test would be nice :)

[xrmx]

Maybe this is enough?

query.name.encode('ascii', 'ignore')

[rumbin]

For a German there is a big difference, whether the Umlauts are entirely dropped or just replaced by their closest ascii counterpart.

I do understand that it may seem inappropriate to introduce yet another dependency for such a minor change, but I feel like this is affecting the user experience, especially of the less technically versed users.

You may safely skip the following pleadings…

---

There are languages, where this is even more relevant. To name an extreme example, in Turkish there exists this nice word düsündürttürücülügümüzün. Personally I think that dusundurtturuculugumuzun may still be readable, but dsndrttrclgmzn is not.

Also think of the Finnish people: A SQLLab tab on the unemployment in the town of Järvenpää työttömyys_Järvenpää could either be tyttmyys_Jrvenp or tyottomyys_Jarvenpaa.

And finally, also our Frensh speaking friends may benefit from not dropping the letters: élevé becomes lev or eleve.

---

[rumbin]

@xrmx: Judging from your smiley, I assume that you expect me to still not having the slightest idea of how to write a test for the csv function that this PR is all about. And you are right.

If somebody (you?) could give me a hand, especially for getting #3705 through, I would be really glad...

[xrmx]

The smile was me begging for tests. Anyway look at tests.core_tests.CoreTests.test_csv_endpoint copy tht test to something like test_csv_endpoint_works_returns_ascii_headers and test that you can decode('ascii') the header.

[mistercrunch]

Unidecode will be useful in other contexts, I'm ok with the new dep.