Improve RepresenterError creation

[bondarevts]

Passing data as an parameter for RepresenterError allows user to make some research about nature of a failed object.

Before:

>>> import yaml
>>> import numpy as np
>>> lst = list(map(np.int16, range(5, 10)))
>>> lst
[5, 6, 7, 8, 9]
>>> try:
...     yaml.dump(lst, Dumper=yaml.SafeDumper)
... except yaml.representer.RepresenterError as e:
...     print(e)
cannot represent an object: 5

Nothing we can do to understand why 5 can't be represented.

After:

>>> try:
...     yaml.dump(lst, Dumper=yaml.SafeDumper)
... except yaml.representer.RepresenterError as e:
...     print(e)
...     if len(e.args) > 1:
...         bad_object = e.args[1]
...         print("Can't represent an object %r of type %s" % (bad_object, type(bad_object)))
...
('cannot represent an object', 5)
Can't represent an object 5 of type <class 'numpy.int16'>

It's even better to have an Exception class with explicit argument for bad object, but I decided to start with smaller changes.

[sigmavirus24]

This is excellent @bondarevts! Thanks!

[sigmavirus24]

This was rebased and merged in ef744d8. Thanks again!