Improve error handling for Image, Images, PreImage, PreImages

[ssiccha]

..., PreImage, and PreImages.

If one of these functions figures out that the second argument is not an
element or subset of the Range or the Source, it now tells the user.

[coveralls]

Coverage increased (+0.007%) to 85.266% when pulling b4ab5d0 on ssiccha:ss/improve-error-handling-image-preimage into 597c793 on gap-system:master.

[codecov]

Codecov Report

Merging #3454 into master will increase coverage by 4.88%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #3454      +/-   ##
==========================================
+ Coverage   80.47%   85.35%   +4.88%     
==========================================
  Files         696      699       +3     
  Lines      343497   346557    +3060     
==========================================
+ Hits       276413   295797   +19384     
+ Misses      67084    50760   -16324

Impacted Files	Coverage Δ
lib/mapping.gi	90.98% <100%> (+1.82%)	⬆️
hpcgap/lib/mapping.gi	90.41% <100%> (+1.21%)	⬆️
lib/gprdpc.gi	45.74% <0%> (-42.2%)	⬇️
src/julia_gc.c	77.4% <0%> (-5.33%)	⬇️
lib/domain.gi	87.69% <0%> (-3.04%)	⬇️
lib/gpprmsya.gi	78.29% <0%> (-2.69%)	⬇️
src/listfunc.c	94.7% <0%> (-1.02%)	⬇️
lib/csetgrp.gi	74.19% <0%> (-0.68%)	⬇️
lib/ghompcgs.gi	91.35% <0%> (-0.67%)	⬇️
src/records.c	87.43% <0%> (-0.66%)	⬇️
... and 192 more

[ChrisJefferson]

Could you write tests to hit all these new cases, just to sanity check them?

[fingolfin]

I am not convinced this is correct, see my inline comment. I guess my claim on what worked before and now doesn't work should be verified. Also, as @ChrisJefferson suggested, tests would be really nice.

[ssiccha]

I'll definitely set up some tests as @ChrisJefferson suggested and have a closer look at your inline comment @fingolfin.

[ssiccha]

I'm updating the PR right now. I have added tests and deleted the

      else
          ErrorNoReturn( "<coll> must be a domain, a set, ",
                         "or a homogeneous list" );

lines. This "else" branch can't be executed since the function compares the families of the source and of the argument elm. The source must be a domain and thus will always have a different family in the situation that elm is not a homogeneous list.

[ssiccha]

PR is updated.

[ssiccha]

And again some HPCGAP tests broke. Let's see whether it works now.

[fingolfin]

In principle this seem to be on a good way, just some minor nitpicking

[ssiccha]

I've addressed your remarks.

[wilfwilson]

I have some requests concerning the error messages, particular about the use of the name <coll>.

[ssiccha]

I adopted the suggested improvements.

I put the changes into a separate commit so you can see more easily what I've changed. I'll turn the new commits into a single one after the PR is accepted.

[ssiccha]

.. and I rebased onto master.

[wilfwilson]

Thanks for your changes.

[ssiccha]

Squashed.

[fingolfin]

Thank you!