Continue porting period_helper to cython

[jbrockmendel]

This moves functions to cython but retains their C names to make comparison easier. The next pass will get rid of the camelCase. A couple of redundant functions are left behind in period_helper, will be cleaned up in a follow-up.

The next hurdle is to get rid of all the adding and subtracting of ORD_OFFSET, WEEK_OFFSET, BDAY_OFFSET. (The period_helper code puts 0 at Jan 1 0001 instead of Jan 1 1970). This would be straightforward but I want to be careful about C division/mod conventions.

[codecov]

Codecov Report

Merging #19608 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #19608      +/-   ##
==========================================
- Coverage   91.62%   91.61%   -0.01%     
==========================================
  Files         150      150              
  Lines       48790    48790              
==========================================
- Hits        44703    44701       -2     
- Misses       4087     4089       +2

Flag	Coverage Δ
#multiple	89.99% <ø> (-0.01%)	⬇️
#single	41.73% <ø> (ø)	⬆️

Impacted Files	Coverage Δ
pandas/util/testing.py	83.64% <0%> (-0.21%)	⬇️

---

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 b835127...769feee. Read the comment docs.

[jreback]

can you run perf on periods

[jreback]

ideally add some full doc-strings on these which list the args and types

[jbrockmendel]

Will do. Mind doing those in the next round to avoid clogging the CI?

[jreback]

that's fine

[jbrockmendel]

taskset 5 asv continuous -f 1.1 -E virtualenv master HEAD -b period -b period -b timeseries
[...]
   before     after       ratio
  [b8351277] [769feee1]
+   19.37μs    23.65μs      1.22  period.PeriodUnaryMethods.time_now('M')
+    2.01ms     2.40ms      1.19  timeseries.ResampleSeries.time_resample('period', '5min', 'mean')
+  133.42ms   154.58ms      1.16  timeseries.DatetimeIndex.time_to_pydatetime('tz_aware')
+    7.40μs     8.44μs      1.14  timeseries.AsOf.time_asof_single_early('Series')
+    2.38ms     2.71ms      1.14  timeseries.ResampleSeries.time_resample('period', '5min', 'ohlc')
+    4.27μs     4.79μs      1.12  timeseries.DatetimeIndex.time_get('dst')
+    2.46ms     2.76ms      1.12  timeseries.ToDatetimeCache.time_dup_seconds_and_unit(False)
-    4.53μs     4.04μs      0.89  timeseries.DatetimeIndex.time_get('tz_naive')
-   71.76μs    63.44μs      0.88  period.PeriodProperties.time_property('M', 'start_time')
-    3.16μs     2.72μs      0.86  timeseries.DatetimeIndex.time_get('repeated')

    before     after       ratio
  [b8351277] [769feee1]
+   20.97ms    25.91ms      1.24  timeseries.IrregularOps.time_add
+  886.50ns     1.01μs      1.14  period.PeriodProperties.time_property('min', 'dayofweek')
+   18.87μs    20.96μs      1.11  period.PeriodUnaryMethods.time_asfreq('M')
-   19.24μs    17.26μs      0.90  timeseries.AsOf.time_asof_single('Series')
-    4.06ms     3.64ms      0.90  timeseries.ToDatetimeISO8601.time_iso8601_nosep
-  962.36ns   852.72ns      0.89  period.PeriodProperties.time_property('min', 'dayofyear')
-  153.68ms   128.59ms      0.84  timeseries.DatetimeIndex.time_to_pydatetime('tz_aware')
-   24.75μs    20.10μs      0.81  period.PeriodUnaryMethods.time_now('M')