low BGS and MWS observation efficiency

I'm not sure whether the underlying problem is in desitarget, fiberassign, or quicksurvey, so assigning this ticket to quicksurvey_example to track it down:

As noted in desihub/desitarget#106, the BGS and MWS fiber assignment efficiency is very low using the example tiles and epochs here. Figure out why and fix it. A snippet of code as a starting point:
```python
from desitarget import desi_mask
from astropy.table import Table, join
targets = Table.read('input/targets.fits')
zcat = Table.read('output/4/zcat.fits')
ztarget = join(targets, zcat, keys='TARGETID', join_type='outer')

#- Trim edge effects
ii = (201 < ztarget['RA']) & (ztarget['RA'] < 209)
ii &= (-7 < ztarget['DEC']) & (ztarget['DEC'] < 7)
ztarget = ztarget[ii]

obs = ~ztarget['Z'].mask
zgood = (ztarget['ZWARN'] == 0)

plot(ztarget['RA'][obs], ztarget['DEC'][obs], 'b,', alpha=0.2)
plot(ztarget['RA'][~obs], ztarget['DEC'][~obs], 'r,', alpha=0.2)

print('Target | Obs Eff | Z Eff')
print('--- | --- | ---')
for bit in range(63):
jj = (ztarget['DESI_TARGET'] & 2**bit) != 0
if np.any(jj):
ntot = np.count_nonzero(jj)
nobs = np.count_nonzero(jj & obs)
ngood = np.count_nonzero(jj & obs & zgood)
obseff = nobs / ntot
zeff = ngood / nobs
print('{:8s} | {:.3f} | {:.3f}'.format(desi_mask.bitname(bit), obseff, zeff))
```

Target | Obs Eff | Z Eff
--- | --- | ---
LRG | 0.950 | 0.967
ELG | 0.779 | 0.977
QSO | 0.859 | 0.941
BGS_ANY | 0.166 | 0.978
MWS_ANY | 0.140 | 0.981