Change hijri calendar dependency to support Python 3

[noviluni]

The current package we are using to parse dates from the Hijri calendar (https://pypi.org/project/ummalqura/) is not maintained and it doesn't support Python 3.

I think that we can use this other package, as it has been maintained and it works pretty well, containing all the features we need: https://github.com/dralshehri/hijri-converter

You can see in the diff below how simple it is to adapt this new package to the current dateparser status (tests are passing).

This library only supports Python 3.6+, but that's not a problem, as we will remove Python 2.7 support in the next release and the end of life of Python 3.5 is 2020-09-13 (the next release of dateparser is planned for September or October, we don't need to remove support for python 3.5 but just add a note about the hijri calendar compatibility).

fixes: #385

[codecov]

Codecov Report

Merging #718 into master will increase coverage by 0.12%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #718      +/-   ##
==========================================
+ Coverage   95.49%   95.61%   +0.12%     
==========================================
  Files         304      304              
  Lines        2662     2646      -16     
==========================================
- Hits         2542     2530      -12     
+ Misses        120      116       -4     

Impacted Files	Coverage Δ
dateparser/calendars/hijri_parser.py	89.74% <100.00%> (+2.24%)	⬆️
dateparser/parser.py	99.02% <0.00%> (+<0.01%)	⬆️
dateparser/date.py	99.57% <0.00%> (+0.80%)	⬆️

---

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 b256f14...97b7335. Read the comment docs.

[Gallaecio]

Looks good to me.

On a separate note, given the next release will remove Python 2 support, is this really a breaking change?

[noviluni]

Oh, it wouldn't be a breaking change.

I am waiting 2 weeks to ensure that the last version is stable enough before proceeding with merging those PRs that would break the Python 2.7 code and marked them with the "breaking change" label to be easily recognizable. I just wrote about this here: #727 (comment)

I agree with you that in the next release notes this shouldn't be labeled as a "breaking change".

[noviluni]

code rebased against the master branch to fix conflicts