dials.merge: Fix output of SigF, N+, N- in merged.mtz

[jbeilstenedmands]

For anomalous=True (the default), when merging anomalous Fs to normal Fs,
don't use_internal_variances. This gives an incorrect (overly high) value for SigF by several factors.

Also, the miller index multiplicities were being written as N+,
N- however this was intended to be the redundancy of observations
of each symmetry group, which is fixed here.

[jbeilstenedmands]

Note: this does not affect the xia2 pipeline which uses cctbx_FrenchWilson to convert Is to Fs, only manual merging with dials.merge

[codecov]

Codecov Report

Merging #1522 (42394cb) into master (fcfd7b4) will increase coverage by 0.02%.
The diff coverage is 94.44%.

@@            Coverage Diff             @@
##           master    #1522      +/-   ##
==========================================
+ Coverage   65.62%   65.65%   +0.02%     
==========================================
  Files         614      614              
  Lines       68964    69059      +95     
  Branches     9522     9559      +37     
==========================================
+ Hits        45256    45339      +83     
- Misses      21868    21879      +11     
- Partials     1840     1841       +1     

[rjgildea]

I think the term "multiplicity" is generally preferable to "redundancy" - can we use this instead? It would also be more consistent with existing usage within dials:

$ find . -name "*.py" -type f | xargs grep -i multiplicit | wc -l
      93
$ find . -name "*.py" -type f | xargs grep -i redundanc | wc -l  
       7

and xia2:

$ find . -name "*.py" -type f | xargs grep -i multiplicit | wc -l
     101
$ find . -name "*.py" -type f | xargs grep -i redundanc | wc -l 
      12