Fix .abstract for classes named with leading underscore(s)

[ceball]

Fix the (longstanding) assumption in param's abstract class mechanism that abstract classes don't start with a leading underscore.

Any identifier of the form __spam (at least two leading underscores, at most one trailing underscore) is textually replaced with _classname__spam, where classname is the current class name with leading underscore(s) stripped. This mangling is done without regard to the syntactic position of the identifier, as long as it occurs within the definition of a class.

(https://docs.python.org/2/tutorial/classes.html)

param/param/parameterized.py

Lines 2074 to 2094 in 301b961

	# Python 2.6 added abstract base classes; see
	# https://github.com/ioam/param/issues/84
	def __is_abstract(mcs):
	"""
	Return True if the class has an attribute __abstract set to True.
	Subclasses will return False unless they themselves have
	__abstract set to true. This mechanism allows a class to
	declare itself to be abstract (e.g. to avoid it being offered
	as an option in a GUI), without the "abstract" property being
	inherited by its subclasses (at least one of which is
	presumably not abstract).
	"""
	# Can't just do ".__abstract", because that is mangled to
	# _ParameterizedMetaclass__abstract before running, but
	# the actual class object will have an attribute
	# _ClassName__abstract. So, we have to mangle it ourselves at
	# runtime.
	try:
	return getattr(mcs,'_%s__abstract'%mcs.__name__)
	except AttributeError:
	return False

(i.e. param does not do the same thing as python)

(originally from #423 (comment))

Also added note about API0 and API1 tests (although not going so far as to say when you should be adding to one or the other or both, just noting the current duplication...)

[coveralls]

Coverage remained the same at 79.447% when pulling 8c1b811 on abstract_fix into 301b961 on master.

[philippjfr]

Looks good to me!

[jbednar]

Looks good, thanks!