Fix/split at when split_point is not in series

[DavidKleindienst]

Fixes #1312

Summary

This PR changes the behavior of TimeSeries._split_at - and therefore also split_before and split_after - when split_point is a Timestamp which is not an index of the TimeSeries but lies between two indeces.
It also adds an appropriate test to the ´TimeSeriesTestCase.test_split` unit test.

Let's say we have a TimeSeries indexed by dates 2020-01-01, 2020-01-03 and 2020-01-05:

Current behavior:
TimeSeries._split_at(2020-01-02, after=True) produces two TimeSeries with indeces [2020-01-01, 2020-01-03] for the first, and 2020-01-05 for the second.
TimeSeries._split_at(2020-01-02, after=False) tries to produce two TimeSeries with indeces [ ] for the first, and [2020-01-01, 2020-01-03, 2020-01-05] for the second, but errors because the first one is empty.

Proposed behavior:
TimeSeries._split_at(2020-01-02) produces two TimeSeries with indeces 2020-01-01 for the first, and [2020-01-03, 2020-01-05] irrespective of whether after is True or False.

I believe this is the behavior most people would expect when splitting a TimeSeries at a point that lies between two dates.

This PR does not change the behavior of TimeSeries splits when split_point is not a Timestamp, or split_point is a Timestamp contained in the series.

Other Information

[codecov-commenter]

Codecov Report

Base: 93.68% // Head: 93.68% // Increases project coverage by +0.00% 🎉

Coverage data is based on head (5f49ed2) compared to base (bcf9abf).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1415   +/-   ##
=======================================
  Coverage   93.68%   93.68%           
=======================================
  Files          94       94           
  Lines        9408     9395   -13     
=======================================
- Hits         8814     8802   -12     
+ Misses        594      593    -1     

Impacted Files	Coverage Δ
darts/timeseries.py	91.85% <100.00%> (+0.02%)	⬆️
...arts/models/forecasting/torch_forecasting_model.py	89.50% <0.00%> (-0.05%)	⬇️
darts/models/forecasting/block_rnn_model.py	98.24% <0.00%> (-0.04%)	⬇️
darts/models/forecasting/nhits.py	99.27% <0.00%> (-0.01%)	⬇️
darts/datasets/__init__.py	100.00% <0.00%> (ø)

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

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[hrzn]

Thanks @DavidKleindienst , this makes sense to me