Fix some pprint_thing warts

[ghost]

Part 1
currently:

In [4]: from pandas.core.common import pprint_thing as pp_t
   ...: pp_t([1,None])
Out[4]: u'[1, ]'

after PR:

In [13]: from pandas.core.common import pprint_thing as pp_t
    ...: pp_t([1,None])
Out[13]: u'[1, None]'

and jeff fixed index NaN printing in #3034, at a higher level.

Part 2

currently:

In [1]: pd.Index(['a','b'])
   ...: 
Out[1]: Index([a, b], dtype=object)

after PR:

pd.Index(['a','b'])
Out[3]: Index([u'a', u'b'], dtype=object)

so:

In [7]: from pandas import Index
In [10]: eval(repr(pd.Index([u'a', u'b'], dtype=object)))
Out[10]: Index([u'a', u'b'], dtype=object)

Which is what repr() should enable, in principle.

[ghost]

added notes to 0.11.0., but maybe this should wait until 0.12/0.11.1

also, make Timstemp repr executable, like datetime.

[jreback]

@y-p I only found the 1 case where a NaN was actually passed to pprint_thing (in index.py/format, which I fixed)

[ghost]

yep. what bothers you about these changes?

[jreback]

is there a test case where Part 1 actually occurs?

[ghost]

say there isn't. Why should None be invisible, even in theory?

[jreback]

oh...nothing....they look fine...just curios (I have always though we should ALWAYS convert None anyhow...but that is a separate issue)

[ghost]

I must have written that at the spur of some 4am moment, "that's no basis for a system of government".

[jreback]

that's pretty funny!

[ghost]

rebase fail, redone with 4aea3f2