Setting quantize default is applied automatically to non-float columns

[CarstVaartjes]

Hi,

i'm fiddling around a bit with chunklen, compression etc. As we have a lot of non-homogenous ctables (mixes of int64 and float64), some of the ctable functionality runs into issues. So if you do this:

bcolz.cparams.setdefaults(cname='zstd', clevel=5, shuffle=2, quantize=16)

you get the following error with a non-homogenous ctabel creation using a fromiter:

/srv/python/venv/local/lib/python2.7/site-packages/bcolz/toplevel.pyc in fromiter(iterable, dtype, count, **kwargs)
    207             # Iterable has been exhausted
    208             break
--> 209         obj.append(chunk)
    210     obj.flush()
    211     return obj

/srv/python/venv/local/lib/python2.7/site-packages/bcolz/ctable.pyc in append(self, cols)
    427                 column = cols[name]
    428             # Append the values to column
--> 429             self.cols[name].append(column)
    430             if sclist and not hasattr(column, '__len__'):
    431                 clen2 = 1

bcolz/carray_ext.pyx in bcolz.carray_ext.carray.append (bcolz/carray_ext.c:21637)()

bcolz/carray_ext.pyx in bcolz.carray_ext.chunk.__cinit__ (bcolz/carray_ext.c:5278)()

bcolz/carray_ext.pyx in bcolz.carray_ext.chunk.compress_arrdata (bcolz/carray_ext.c:5924)()

/srv/python/venv/local/lib/python2.7/site-packages/bcolz/utils.pyc in quantize(data, significant_digits)
    178 
    179     if data.dtype.kind != 'f':
--> 180         raise TypeError("quantize is meant only for floating point data")
    181 
    182     if not significant_digits:

TypeError: quantize is meant only for floating point data

I will try to be helpful and see if i can see to make this smarter with a PR that will ignoring quantization for non-float columns. Also might create a PR to have a more complex chunklen for ctables (per column), but that's not the issue here

[FrancescAlted]

Yeah, a PR on that would be great. Thanks in advance!