fix(rfi): fix misalignment between mask and data in sumthreshold

[ssiegelx]

Also removes unnecessary slices.

[ssiegelx]

Here is a simple example illustrating what I think is a bug in the sumthreshold method:

nx = 256
data = np.zeros(nx)

m = 32
i = nx // 2

data[i:(i+m)] = 1.0

## Taken from origin sumthreshold
centre = m // 2 - 1

dconv = convolve1d(data, weights=np.ones(m, dtype=float), origin=-centre)

flag_temp = dconv >= m

flag = np.zeros(nx, dtype=bool)
for ii in range(nx - m + 1):
    flag[ii:(ii+m)] += flag_temp[ii]


plt.plot(data, marker='o', color='k')
plt.plot(dconv / m, color='b')
plt.plot(flag, color='r')

You can see that instead of the signal being flagged, the data immediately after the signal is flagged.

Here is the result if we use origin=centre instead:

The mask is properly aligned with the excursion.

[ljgray]

Interesting - is the behaviour the same if m is odd?

[ssiegelx]

For m = 1 it will be properly aligned, but for other odd values of m there will be a shift that is proportional to the size of the window.