In read_csv, setting index_col breaks functionality of converters when engine='python'

[gilwolff]

I am experiencing an issue with read_csv
When setting index_col, and a converters dict, output of engine='c' is correct and output of engine='python' is wrong

def first(s):
    return s[0]
io = StringIO('col1,col2\n1_,a\n2_,b\n')
print pd.read_csv(io, index_col=0, converters = {0 : first})
     col2
col1
1       a
2       b
io = StringIO('col1,col2\n1_,a\n2_,b\n')
print pd.read_csv(io, index_col=0, converters = {0 : first}, engine='python')
     col2
col1
1_      a
2_      b

Expected output: for second example to be like the first
Note: If I leave out index_col=0, converter works as expected

Output of pd.show_versions():

commit: None
python: 2.7.12.final.0
python-bits: 64
OS: Linux
OS-release: 4.4.11-23.53.amzn1.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8

pandas: 0.18.1
nose: None
pip: 8.1.2
setuptools: 12.2
Cython: None
numpy: 1.11.1

[jreback]

this is detailed in #9435, last example by @gfyoung; the logic for handling convererts is a bit convoluted ATM. pull-requests to fix would help.