to_html: sparsify mutliindex keys in dataframe rows

[btel]

The sparsify option of DataFrame.to_html currently works only for columns. It would be useful to add this option also for rows, similarly to what is already done for data frame representation in console.

Example:

df = pd.DataFrame(np.arange(3), index=[['X', 'X', 'Y'], ['a','b','a']])
df.to_html(fid)

currently produces the following output:

		0
X	a	0
X	b	1
Y	a	2

This should be changed to:

		0
X	a	0
	b	1
Y	a	2

[wesm]

Looking pretty good now:

	exp	A	B	C
first	second
foo	one	-1.159321	-0.131931	0.818024
	two	-1.318878	0.285172	0.551480
	three	0.340577	0.376841	0.050759
bar	one	1.300129	-1.733227	-0.746667
	two	0.178495	-0.286725	0.011206
baz	two	0.764214	0.702831	0.080559
	three	-0.299611	-0.505831	-0.319067
qux	one	-0.413753	1.213659	-0.234424
	two	-1.029432	-1.984151	-0.047421
	three	0.645149	-0.683223	-0.544720