Improve performance for TimeSeriesResult.to_df

[joloppo]

This still does not solve threadpool being slow for low dimensionality data. ie only 2 columns and only <100 or so records.

However, this improves one part of the code that was very slow, and disables the theadpool if only 1 core is available.

RELATED TICKETS

https://vortexa.atlassian.net/browse/RND-7233

CHANGELOG

• Disable threadpool if only 1 core available.
• Improve performance of some code.

TESTS

• Leave tests in, possible more perf tests should be done

Perf

Perf testing results for thousands of rows & columns:

1 year worth of data for this:
CargoTimeSeries().search(
            timeseries_frequency='day',
            timeseries_property='origin_terminal',
            timeseries_activity='loading_end',
            filter_activity='oil_on_water_state',
            filter_time_min=datetime(2021, 1, 1),
            filter_time_max=datetime(2021, 12, 31))

127.51475930213928 seconds -> mp pool 
175 seconds -> no mp pool

pool with fix 
(44543, 2571)
Time taken to convert to DataFrame: 38.43424606323242 seconds

nopool with fix
(44543, 2571)
Time taken to convert to DataFrame: 86.82875204086304 seconds