Duplicate rule suppression sometimes fails

As discussed in https://github.com/openwall/john/issues/5004#issuecomment-1019310089:

Shouldn't the current code have rejected a repeated rule like the below (when running -ru=jumbo)?

$ fgrep 'c Az"111"' sim3.log 
0:00:02:49 - Rule #923: '-c (?a c Az"111" <+' accepted as '(?acAz"111"<+'
0:00:02:49 - Score 0.012869018 at 10490.31 p/g 135g 1416192p during rule #923 :-c (?a c Az"111" <+
0:00:08:32 - Rule #2769: '-c (?a c Az"111" <+' accepted as '(?acAz"111"<+'
0:00:08:33 - Score 0.000000001 at 1416192000000000.00 p/g 0g 1416192p during rule #2769 :-c (?a c Az"111" <+

Obviously, it doesn't crack anything the second time this rule is tried. The log has this near the start:

0:00:00:00 - 72076 preprocessed word mangling rules were reduced by dropping 11 rules
0:00:00:00 - 72065 preprocessed word mangling rules

We should find out why the duplicate rule wasn't dropped.

Start by reproducing the -ru=jumbo run and inspect the preprocessing output and sim3.log entries for rule #923 and rule #2769. Trace the duplicate-rule suppression path to determine why these normalized rules are both retained. Done means the repeated rule is dropped during preprocessing and the logs no longer show it being tried twice.