Variables protected by mutex are incorrectly calculated

[sim642]

./regtest.sh 46 98 --enable dbg.print_protection

outputs

[Info][Race] Variable nothing2 read-write protected by 1 mutex(es): {f}
[Info][Race] Mutex f read-write protects 1 variable(s): {nothing2}
[Info][Race] Variable j read-write protected by 1 mutex(es): {f}

The first and third lines use the dynamically computed "mutexes protecting the variable" information: both are protected by f.
The second line uses the inverse information which is only computed during postsolving, but somehow it only lists nothing2, but should also list j.

I think this information is only used for dbg.print_protection and witness-generation, which is probably why it has gone unnoticed for so long.

PR #1631 identified that write:bool is used with different meanings but didn't reveal any issues arising from it, only confusion. I suspect this bug might actually be caused by the mixup.

[sim642]

I now tried #1631 and it seems to have been (accidentally?) fixed there:

[Info][Race] Variable nothing2 read-write protected by 1 mutex(es): {f}
[Info][Race] Variable j read-write protected by 1 mutex(es): {f}
[Info][Race] Mutex f read-write protects 2 variable(s): {j, nothing2}