Reduce memory usage of tests

[ChrisJefferson]

This patch unbinds variables in some tests which use a lot of memory.

I built this by running each test and measuring memory usage before and after, and unbinding in the worst ones. Long term, #1633 would fix this general problem, but that really needs some care before merging (in particular, if someone wants to figure out a better way of making a bag of local variables other than my horrible eval method).

[codecov]

Codecov Report

Merging #1791 into master will increase coverage by 1.22%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1791      +/-   ##
==========================================
+ Coverage    61.6%   62.82%   +1.22%     
==========================================
  Files         968      969       +1     
  Lines      295076   295193     +117     
  Branches    13049    13050       +1     
==========================================
+ Hits       181774   185454    +3680     
+ Misses     110522   106940    -3582     
- Partials     2780     2799      +19

Impacted Files	Coverage Δ
src/hpc/traverse.c	77.9% <0%> (-0.7%)	⬇️
src/gasman.c	82.65% <0%> (-0.31%)	⬇️
hpcgap/lib/hpc/stdtasks.g	38.61% <0%> (-0.26%)	⬇️
src/gap.c	56.75% <0%> (-0.09%)	⬇️
lib/debug.g	10.18% <0%> (ø)
lib/oprtperm.gi	92.12% <0%> (+0.11%)	⬆️
src/c_methsel1.c	67.61% <0%> (+0.13%)	⬆️
lib/tom.gi	54.31% <0%> (+0.14%)	⬆️
lib/alghom.gi	55.48% <0%> (+0.16%)	⬆️
lib/semirel.gi	70.86% <0%> (+0.16%)	⬆️
... and 158 more

[fingolfin]

This seems sensible. Not "elegant", of course, but pragmatic, esp. if it solves the problems @alex-konovalov is seeing...

[olexandr-konovalov]

Thank you - agree with this. We need to resolve #1556 for GAP 4.9, and this will hopefully do the job.

I was also thinking of a check of package loading - what are the packages that cause the biggest expansion of the workspace.

[ChrisJefferson]

I had a quick look at packages, and while they increase by different amounts at loading, none jumped out at me as out of proportion -- some packages are just big.

I would suggest increasing the memory allowance to say 2g, I don't think that's unreasonable, and will stop tests failing. If we want to track increasing memory usage, we should do that seperately and consistently.

[olexandr-konovalov]

I will merge this and see if this helps to #1556. Will keep in mind what you say for -o 2g for later.