Seperate half life weights into its own function

[jwaterfield]

Remove the half life weights from the fill_spectrum functions and have it in its own function which can be applied seperately.

Edit 15/04/2015:
Jeanne's suggestion is to remove the timing axis from the Spectra class and have the timing applied as analytic function. This means that whenever you return a number of counts from the spectrum i.e. in Spectra.sum() you would need to multiply the number of events from the 2D spectrum by the result of the correct time model.

[arushanova]

I think we shouldn't do it. Realistically the best precision in time we can get counts in months or weeks(?), i.e. I doubt that anyone would want (would be able) to study the changes in spectra within days or hours. So what we can do is to

1. decide what would be a time unit: a week or a month
2. change the number of time bins from 10 years to 10_52.18 weeks or 10_12 months
3. fill the spectrum as we were doing it before

We have a function that slices spectrum, so it will be possible to study any time block.

[jwaterfield]

Timing will now be analytic

[ashleyrback]

I think this issue isn't actually relevant any more, since all spectral parameters are specified via config now. Are we happy to close this issue?