Improve serialization of geopandas objects and code cleaning

[mattijn]

This PR mainly targets the topojson.core.extract.py file.

It refactors the parsing of GeoDataFrames and GeoSeries. Now a GeoDataFrame is serialized from gdf.to_dict(orient='records') and a GeoSeries from series.to_dict(). No need to go through the much slower __geo_interface__.

Also this PR prepend non public attributes and functions in the Extract class with a _.
And because I removed the options object out of the .options, Ihad to change some other classes and tests as well.

Also more docstrings are included.

[mattijn]

A check with the geojson from #73 (comment) (44mb, 3826 geometries) results in:

• a speedup of 14.7X [from 1160ms (±140ms) to 79ms (±1.67ms)]

And for naturalearth_lowres:

• a speedup of 13.4X [from 78.8ms (±1.55ms) to 5.89ms (±0.12ms)]

!Speedup counts only for the Extract class

[mattijn]

As a side effect this PR also introduce support for input objects structured as:

data = [
    {'geometry': <shapely geometry>, 'y': 1, 'z': 2}, 
    {'geometry': <shapely geometry>, 'y': 2, 'z': 3}
]

and

data = {
    0: <shapely geometry>, 
    1: <shapely geometry>
}

[martinfleis]

a speedup of 14.7X [from 1160ms (±140ms) to 79ms (±1.67ms)]

That is great!