Not all finitely generated magmas are groups (alternative to PR #2276)

[fingolfin]

... hence we should not set IsFinitelyGeneratedGroup for them.
Instead, only set IsFinitelyGeneratedGroup if the final magma
is actually a group (i.e., associative).

This is an alternative to PR #2276
Fixes #2252, closes #2276

[james-d-mitchell]

Looks good to me.

[ThomasBreuer]

The code is correct after these changes,
in this sense I could approve them,
as an intermediate solution.
On the other hand, the drawback is the forced
associativity check in the construction.

[codecov]

Codecov Report

Merging #2311 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2311      +/-   ##
==========================================
+ Coverage   72.52%   72.52%   +<.01%     
==========================================
  Files         478      478              
  Lines      246816   246818       +2     
==========================================
+ Hits       179003   179009       +6     
+ Misses      67813    67809       -4

Impacted Files	Coverage Δ
lib/magma.gi	54.59% <100%> (+0.17%)	⬆️
hpcgap/lib/hpc/stdtasks.g	38.61% <0%> (-0.26%)	⬇️
src/c_oper1.c	91.47% <0%> (+0.16%)	⬆️
src/objset.c	81.49% <0%> (+0.22%)	⬆️

[fingolfin]

@ThomasBreuer you are of course right; and this forced check also changes the outputs of various tests (causing them to fail), and indeed can produce a hang if the associativity check is difficult. I have now replaced this by what should have been in the first version of this PR: a check for HasIsAssociative( M ) and IsAssociative( M ) which avoids all that.

Thanks for the careful reviewing, I really appreciate it!