We need to use the correct subject when doing global scanning,
too. In fact, the per-recipient spam training path is entirely
redundant at this point.
my $name = $ibx->{name};
my $addr = $ibx->{-primary_address};
my $im = PublicInbox::Import->new($git, $name, $addr, $ibx);
my $name = $ibx->{name};
my $addr = $ibx->{-primary_address};
my $im = PublicInbox::Import->new($git, $name, $addr, $ibx);
+ $im->remove($mime, 'spam');