Fix/onedimensional sigma

[Moritz-Alexander-Kern]

This PR addresses Issue #455 .

Changes

• check dimensions of sigma, if dimension > 0 , raise typeError
• handle unambiguous case ndim=1 , size = 1 with sigma = sigma[0]
• added tests for both above cases

Test

The following code snippet should no longer throw an error

import neo
import elephant
import quantities as pq

st = neo.SpikeTrain([1]*pq.s, t_stop=2*pq.s)
kernel = elephant.kernels.GaussianKernel(sigma=[20]*pq.ms)

rate = elephant.statistics.instantaneous_rate(st,
                                              sampling_period=1*pq.ms,
                                              kernel=kernel)

[Kleinjohann]

Looks good, thanks for the fix, I can confirm that my use case that originally caused this error is covered by this.

[Kleinjohann]

Just a side remark:
This will also catch something like sigma = [[5]] * pq.s
Seems fine to me though, since a meaningful error is raised in that case now.

[mdenker]

I assume that in the long run -- besides this not being Python -- we may want to have a more standard way to check types, and in particular the case of 0, 1 and 2 dimensional quantities arrays.