validate that SV/BGS targets are receiving the correct (cumulative) number of fibers

Creating this issue from [desi-sv 21](https://desi.lbl.gov/trac/tracmailman/browser/private/desi-sv/2018-July/000020.html) by @kdawson1000, although I think it's closely related to https://github.com/desihub/fiberassign/issues/133.

In essence, we just need to be able to set `NUMOBS` in [`desitarget.targets.calc_numobs`](https://github.com/desihub/desitarget/blob/master/py/desitarget/targets.py#L417).

```
[snip]
What then should we do on the last run of BGS fields? The most obvious
question regards the low number of repeat observations for each target.
It would be best to understand the origin of that problem. As a
reminder, I expected to have a large number of targets getting four
fibers over all of the tiles, but that did not happen. The prediction
based on randoms and a simple 1.6 degree radius focal plane was a
fraction 0.09 0.12 0.30 0.45 of targets would get 1,2,3,4 fibers,
respectively. Instead, the assignment code gave 0.42 0.31 0.17 0.10.

One theory is that sky and std star targets are randomly bumping targets
and thus reducing the average number of fiber assignments per target.
There are other possible reasons. One may want to resolve this by
subsampling the BGS catalog before running fiber assignment, thus
reducing the pressure on sky and std star fibers.
[snip]
```

Start with desitarget/py/desitarget/targets.py at calc_numobs, then review the related fiberassign issue 133 and the quoted SV/BGS assignment results. Determine how cumulative NUMOBS should be represented and how to validate assigned fibers across tiles. Done means the expected and observed fiber counts agree for SV/BGS targets, with coverage for the relevant cases.

Written by the indexing model from the issue text.