Consider to always return datetime-indices

[berland]

get_smry() and load_smry() is not very consistent on what kind of index they return.

Consider going all in and always have the time-series indexed properly with Pandas datetime64 DateTimeIndex.

This would be a breaking change in the API, requires bumping major revision.

[berland]

Using Pandas datetime64[ns] as default is a no-go due to the hard upper limit in year 2262.

The realization object returns summary dataframes with dates in the index (datetime64[ns] or datetime.datetime objects if necessary from ecl2df v0.11). This can continue as in fmu.ensemble 1.4 (it is returning libecl's pandas_frame() directly).

load_smry() in fmu-ensemble 1.4 realization.py performs a reset_index(). This will not be carried over to fmu.ensemble 2.0.

The Ensemble objects cannot have dates in their index, and is always doing a reset_index() (effectively) on dataframes from realization objects). This behaviour should be kept going to fmu.ensemble 2.0

[berland]

Current decision is to have a dummy index all-over fmu-ensemble, and DATE in a separate column. This is an API change from fmu.ensemble 1.5.0 to 2.0.0, in PR #182