Revise photometric fitting

This commit considerably improves the photometric fitting procedure.
The most important changes are:
1) Photometric fitting does now consider quality flags.
1.1) In order to make this work, it was necessary to add the field
'flag' to the photometric data tables created by 'photometric_table'.
This means that this commit is not backward compatible with existing
photometric data files. Just create new ones with 'query_photometry',
which will then also have the correct flags from the catalogs, see 1.2.
1.2) The function 'query_photometry' flags data points according to
the quality criteria given by the individual photometric catalogs.
1.3) If the new qualifier 'remove_outliers' is used, the function
'photometric_fitting' identifies outliers according to the worst
offender principle and flags them, which means that those data points
are then ignored in the subsequent fitting procedure.
2) It is no longer necessary to re-compute the grid after new photo-
metric filters have been added. Existing photometric magnitudes are
now read from the grid while missing ones will be computed on the
fly. Because on-the-fly computations take considerably longer than
just reading them, performance will be much better when all magnitudes
are present in the grid. The function 'photometric_fitting' will print
a message if on-the-fly computations are necessary. Three box filters
are now added to the template 'make_fit_grid.sl' to improve the
performance when box filters created from IUE spectra are fitted.
3) Because the fit algorithm "powell" has serious problems with small
parameter values, the fitting parameter "theta" has been changed to
"logtheta", i.e., log10(theta), in all fit-functions. This means that
old parameter files containing the fit-parameter "theta" cannot be
loaded any longer. This also applies to spectroscopy when flux-
calibrated spectra were fitted. To solve this, remove the parameter
file and redo the fit from scratch or change the file by hand, i.e.,
rename the parameter "theta" to "logtheta", change its value as well
as its min- and max-values.
4) The treatment of box filters from IUE spectra has been improved.
Spectra are now taken from INES ("IUE Newly Extracted Spectra",
http://sdc.cab.inta-csic.es/ines/). Moreover, statistical uncertainties
- based on propagating the uncertainties in the flux to uncertainties in
the box filter assuming Gaussian error propagation - are now assigned
to each individual magnitude. If box filter magnitudes from several
IUE spectra are available, they are averaged and the respective
standard deviation is assigned as uncertainty. This averaged magnitude
is then by default used in the fitting process. The individual magnitudes
are flagged as redundant.
5) The function 'xfig_photometry' now features the new qualifier 'chi'.
If set, residual panels for magnitudes and colors will show the difference
between model and observation divided by the uncertainty instead of only
the difference. This is useful because the chi-values (and not the
differences) are minimized during the fitting procedure. Moreover,
chi-values have a typical parameter range making it easy to set
proper ticmarks for the axes, which frequently failed so far.
6) Finally, two new functions were added which make it very easy to
collect and TeX photometric results: 'photometry_collect_results'
and 'photometry_TeX_table'.