Adding support of pd.RangeIndex for Series and DFs

[kozlov-alexey]

This PR:

1. Modifies boxing/unboxing of Series and DFs to handle pd.RangeIndex,
2. Adds fix_df_index to transform values of index argumenent of Series and
   DF ctor calls, which fixes RewriteDataFrame ctor now handling index=None as argument,
3. Adds iteration, operators (is, eq, ne) support for RangeIndexType,
4. Renames and refactors sdc_check_indexes_equal to numpy_like.array_equal,
5. Adds specializations for RangeIndexType in all Series/DF methods,
   such as operators, getitem, setitem and indexing related functions
   (sdc_join_series_indexes, sdc_reindex_series, etc).

[pep8speaks]

Hello @kozlov-alexey! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2020-06-24 12:21:45 UTC

[kozlov-alexey]

It seems performance degraded with these changes, need to fix that before merge, e.g. for
operator.add:

with #862:

1	Python	10000000	0.031289	0.03108	0.033158
1	SDC	10000000	0.188526	0.188433	0.188893	2.371142	0.000554
2	SDC	10000000	0.152458	0.152053	0.153302	2.43293	0.000682
4	SDC	10000000	0.086953	0.086793	0.087348	2.425093	0.000674
8	SDC	10000000	0.053097	0.053085	0.053173	2.430201	0.000825
16	SDC	10000000	0.036783	0.036191	0.037241	2.418686	0.000837
28	SDC	10000000	0.03135	0.031309	0.031361	2.425914	0.000833
56	SDC	10000000	0.029505	0.02948	0.030095	2.413604	0.000696

on master:

1	Python	10000000	0.03126	0.031004	0.031958
1	SDC	10000000	0.047348	0.047187	0.047402	0.393299	0.00041
2	SDC	10000000	0.033103	0.033065	0.033186	0.446369	0.000514
4	SDC	10000000	0.017455	0.017369	0.017512	0.44836	0.000512
8	SDC	10000000	0.008907	0.008868	0.010109	0.444994	0.000514
16	SDC	10000000	0.005434	0.005059	0.007267	0.443173	0.000626
28	SDC	10000000	0.004042	0.003909	0.005736	0.444903	0.000519
56	SDC	10000000	0.004228	0.004043	0.006067	0.445996	0.000637

[kozlov-alexey]

With reverting back to a separate impl for none (positional) indexes perf stays the same with this PR, see below for Series.operator.add:
with this PR:

nthreads	type	size	median	min	max	compile	boxing
1	Python	10000000	0.031252	0.031215	0.031294
1	SDC	10000000	0.049261	0.049189	0.049536	0.4135	0.000481
2	SDC	10000000	0.033909	0.033767	0.033971	0.448139	0.000529
4	SDC	10000000	0.017572	0.017527	0.017576	0.453382	0.000608
8	SDC	10000000	0.00911	0.009097	0.009185	0.446261	0.000554
16	SDC	10000000	0.005	0.00499	0.005736	0.465148	0.000619
28	SDC	10000000	0.003969	0.003954	0.005549	0.454575	0.000649
56	SDC	10000000	0.004146	0.00411	0.004553	0.451319	0.000679

on master (e061965):

nthreads	type	size	median	min	max	compile	boxing
1	Python	10000000	0.031025	0.030967	0.031119
1	SDC	10000000	0.049254	0.049175	0.049279	0.392598	0.000421
2	SDC	10000000	0.034642	0.03444	0.034732	0.450111	0.000578
4	SDC	10000000	0.017848	0.017798	0.018241	0.45714	0.000616
8	SDC	10000000	0.009329	0.009308	0.010389	0.452543	0.000596
16	SDC	10000000	0.005169	0.005132	0.007395	0.448971	0.000621
28	SDC	10000000	0.004017	0.003881	0.005692	0.440998	0.000525
56	SDC	10000000	0.004092	0.004059	0.006165	0.462535	0.000664