Fix for computing quotients of certain algebra modules

[fingolfin]

The issue was originally reported by Rudolf Tange on the GAP support mailing list.

[fingolfin]

This is the error on HPC-GAP:

gap> V:=DirectSumOfAlgebraModules(s);
# Expected output:
<35-dimensional left-module over <Lie algebra over GF(5), with 3 generators>>
# But found:
Error, Atomic List Element: <pos>=2 is an invalid index for <list>

If somebody wants to look into it, be my guest...

[fingolfin]

There was an alternative fix suggested by Willem De Graaf on the mailing list. At least one of them should be in GAP 4.9.

[markuspf]

This needs fixing for HPC-GAP (and probably a rebase).

[olexandr-konovalov]

For the record, Willem's reply is in his email "[GAP Support] BUG in DirectSumOfAlgebraModules or inquotient-module function?" on 14/09/2017.

[codecov]

Codecov Report

Merging #1669 into master will increase coverage by 0.07%.
The diff coverage is 46.15%.

@@            Coverage Diff             @@
##           master    #1669      +/-   ##
==========================================
+ Coverage   63.49%   63.56%   +0.07%     
==========================================
  Files         952      952              
  Lines      286804   286813       +9     
  Branches    12722    12722              
==========================================
+ Hits       182103   182311     +208     
+ Misses     101899   101708     -191     
+ Partials     2802     2794       -8

Impacted Files	Coverage Δ
hpcgap/lib/basis.gi	79.23% <33.33%> (-0.27%)	⬇️
lib/basis.gi	80.42% <33.33%> (-0.28%)	⬇️
lib/algrep.gi	69.14% <57.14%> (+12.1%)	⬆️
src/funcs.c	74.61% <0%> (-0.15%)	⬇️
src/opers.c	78.48% <0%> (+0.05%)	⬆️
src/listoper.c	75.44% <0%> (+0.1%)	⬆️
src/lists.c	56.71% <0%> (+0.11%)	⬆️
src/stats.c	79.15% <0%> (+0.13%)	⬆️
src/hpc/thread.c	45.65% <0%> (+0.19%)	⬆️
src/hpc/threadapi.c	34.49% <0%> (+0.28%)	⬆️
... and 3 more

[fingolfin]

I updated this PR to use the fix proposed by Willem instead of my original fix, as his fix corrects additional problems -- one is that IsCanonicalBasis did not terminate if called on a basis of a sparse row spaces; that problem persisted with my fix, but is corrected by Willem's more fundamental fix.

I also tracked down and corrected the problem on HPC-GAP, and added another commit which very slightly improves performance for \in on enumerators-by-basis.