Change kgtkreader and kgtkwriter to process Pandas dataframes

[CraigMiloRogers]

To better integrate with the Python Notebook environment, change kgtkreader and kgtkwriter to process Pandas dataframes directly.

[CraigMiloRogers]

Adding this support at such a low level will help make it pervasive, but will make kgtk be dependent upon pandas in a more fundamental way than it is at present.

[CraigMiloRogers]

Alternatively, we can handle dataframe processing in the api wrappers, much as it is treated in functions.py, but that might be a bit verbose.

[CraigMiloRogers]

Changes committed to dev.

[CraigMiloRogers]

Need an option to convert KGTK strings to native datatypes (int, float, date/time?) when writing a DataFrame.

[CraigMiloRogers]

Need Documentation.

[CraigMiloRogers]

There are now two dataframe output formats for KgtkWriter.

• OUTPUT_FORMAT_DATAFRAME_STRING writes a dataframe containing KGTK-formatted string values. It is fast and no information is lost.
• OUTPUT_FORMAT_DATAFRAME_NATIVE writes a dataframe with Python strings, ints, floats, and bools. It is slower, information is lost, and the conversion is incomplete (e.g., dates could be converted to a native type, but are not yet done).