Add verbosity parameter

[rhiever]

Allow the user to tune the level of output from TPOT. Default to no output.

[MichaelMarkieta]

I'm not seeing any output in python3 when running fit for any verbosity level. What is one of the first things TPOT is supposed to report, and how soon after starting fit should that happen? This is running in the interpreter.

[MichaelMarkieta]

Looks like there is very little in the way of print statements... only at the end of the fit routine do we learn the best pipeline. I thought there would be more along the way. Hmm.

[rhiever]

With maximum verbosity (=2), it should also report the best/worst/average scores every generation. Depending on your data size, each generation may take a while.

We've been thinking about ways to provide an update on the best model every generation in #2. What information are you looking for while TPOT is running?

[MichaelMarkieta]

I got greedy... the dataset was too large and I didn't wait long enough for the first gen (0) to get reported. Working as expected. Carry on...

I am used to the verbosity of http://h2o.ai which is an unfair bias.

[rhiever]

It would probably help if we added an "Expectations" section explaining that TPOT will be rather slow for large data sets. Although in #59 we're exploring ideas on speeding the initial seeding phase up to get the user something useful right off the bat.

[MichaelMarkieta]

Could also print some stuff out right off the bat to tell the user things have been initialized. The fit statement is lonely waiting for gen 0 to finish.

[rhiever]

True enough. We do this for the command-line version, so it seems reasonable to do it for the scripted version as well. Filed as issue #74.