remove_accents: refactor to NFKD normalization

[bosd]

Refactor the remove accents functions to NFKD normalization form. So référence get normalized to reference.

This fixes: #444

[rmilecki]

Looks good, could we also add some test for that?

[bosd]

Looks good, could we also add some test for that?

I agree that it is advisable to cover it in tests.

My preference is to fasttrack this one. Generate a new release on pypi. As there are some minor bugs in the current release.

Then in a separe PR create more advanced tests for the normalize function.
(I am propably not the one to write these advanced tests because we don't have it in my language.)
I could imagine someone wants to add test for german umlaut (was mentioned somewhere here)
As well as the polish & french language might hase some characters which might be affected by this..

@rmilecki If you are ok, with fasttracking this one in the current form w/o tests. Please merge this one 👍

[bosd]

Sadly, This is not a correct solution. It now also deletes the € sign.