Support contigous vector of doubles in ecl_sum_data_add_new_tstep()

For speed reasons, there should be a version of `ecl_sum_data_add_new_tstep()` that can take a contiguous vector of doubles.

The current code takes one value at a time, and performs index buiding each time. This results in slowness when f.ex. EclSum objects are created from numpy vectors/pandas dataframes.

https://github.com/equinor/ecl/blob/265dd23808f3987c38802b018f361b2fe0f317a6/lib/ecl/ecl_sum_data.cpp#L673-L678

Fixing this might create a possibility for test code that creates EclSum objects on the fly.