Replaced OrderedDict with dict

[smithdc1]

dict is now assumed to be ordered in all currently supported versions of Python. This pull request changes OrderedDict to dict across the project. While making the changes I did some tests to show the performance benefit.

1. Preserve order and remove duplicates via fromkeys() (e.g. line 334 of openapi.py)

%timeit list(dict.fromkeys(field.choices))
553 ns ± 1.97 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

%timeit list(OrderedDict.fromkeys(field.choices))
680 ns ± 5.24 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

Conclusion dict is 19% faster

2. Create a dict from a list

(e.g line 221 of pagination.py)

%timeit Response(OrderedDict([('count', self.page.paginator.count), ('next', self.get_next_link()), ('previous', self.get_previous_link()), ('results', data)]))
53.2 µs ± 204 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

%timeit Response({'count': self.page.paginator.count, 'next': self.get_next_link(), 'previous': self.get_previous_link(), 'results': data,})
52.1 µs ± 473 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

(e.g line 660 in renderers.py)

%timeit -r20 OrderedDict(sorted(response.items()))
900 ns ± 9.79 ns per loop (mean ± std. dev. of 20 runs, 1000000 loops each)

%timeit -r20 dict(sorted(response.items()))
684 ns ± 3.59 ns per loop (mean ± std. dev. of 20 runs, 1000000 loops each)

Conclusion dict faster, 2% in the first example, 25% in the second. I suspect in practice the benefit is closer to the first; however, it is consistently faster.

3. Adding key,values to a dict (e.g. line 62 or model_meta.py)

# OrderedDict()
%timeit for field in [field for field in opts.fields if field.serialize and not field.remote_field]: fields[field.name] = field
  ...:     
517 ns ± 4.83 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

# dict
%timeit for field in [field for field in opts.fields if field.serialize and not field.remote_field]: fields[field.name] = field
   ...:     
500 ns ± 4.1 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

Conclusion dict is marginally (3%) faster

[carltongibson]

Django 2.2 still supports Python 3.5.

(But yes... after that.)