Different sublist patterns in ConstructAllGroups outputs

[olexandr-konovalov]

This is meta-issue to collect as much as possible cases where we observe different sublist patterns in ConstructAllGroups outputs. I don't know a reason to doubt replication results, since this is certainly possible when randomised algorithms are being used, so my plan is to add such gnu(n) to the database and close corresponding issues/PRs, but add a link to them to this issue. Such cases may be good candidates for regression tests.

At the moment, we have observed:

• gnu(9792)=73284 #70: 10 pairs of groups in GAP 4.8.3 vs 7 pairs of groups in the master branch for n=9792
• gnu(7290)=3623 #82: no sublists in GAP 4.8.3 and one pair of groups in the master branch for n=7290
• gnu(6804)=5981 #83: 11 pairs of groups in GAP 4.8.3, 8 pairs with GAP 4.8.4 and 10 pairs of groups in the master branch for n=6804
• gnu(7360)=10743 #84: 3 pairs of groups in GAP 4.8.3 vs 4 pairs in GAP 4.8.4 and no sublists in the master branch for n=7360
• gnu(6720)=106765 #99: no sublists in GAP 4.8.3 and 3 pairs of groups in the master branch for n=6720