Set threshold to account for too many duplicate or nan values in DateTimeFormatDataCheck

[ParthivNaresh]

Currently if a time_index is chosen that has duplicate or nan values, these values are dropped prior to the creation of an alias_dict in Woodwork which iterates over all identified frequencies in the time series to determine the most likely one.

The side effect of this is that in a dataset with a length of 20,000, if the time_index is just 50 consecutive daily datetime values repeated 400 times, then we currently return DATETIME_HAS_UNEVEN_INTERVALS and an action code to fix this data.

This time_index should not be considered for regularization as regularizing this data would reduce this dataset from 20,000 observations to 50, since all duplicates and nans would be dropped and the remaining datetime values would have a frequency of 1 day. This is more of a multi-series concern which is outside the scope of this data check.

I've added a check where if 25% of the time_index consists of solely duplicate and nan values, then DATETIME_NO_FREQUENCY_INFERRED will be returned instead of DATETIME_HAS_UNEVEN_INTERVALS. I'm open to changing this threshold or making it a passable argument, it was more or less arbitrary.

[codecov]

Codecov Report

Merging #3883 (5f0cd2d) into main (24a01ae) will increase coverage by 0.1%.
The diff coverage is 100.0%.

@@           Coverage Diff           @@
##            main   #3883     +/-   ##
=======================================
+ Coverage   99.7%   99.7%   +0.1%     
=======================================
  Files        346     346             
  Lines      36284   36304     +20     
=======================================
+ Hits       36147   36167     +20     
  Misses       137     137             

Impacted Files	Coverage Δ
evalml/data_checks/datetime_format_data_check.py	100.0% <100.0%> (ø)
...ta_checks_tests/test_datetime_format_data_check.py	100.0% <100.0%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[ParthivNaresh]

Changing this value crosses the threshold into DATETIME_NO_FREQUENCY_INFERRED

[eccabay]

Since the threshold was arbitrary, I think it certainly wouldn't hurt to make it a passable argument. Regardless, looks great!

[christopherbunn]

Makes sense to me! +1 to @eccabay's comment on making it a passable argument or even just an internal global variable.