Issue Description:

DROOLS-7372 fixed for a bind variable in the same pattern, but a bind variable from another pattern causes the same reactivity issue.

example 1

rule "Rule 1"
    when
        $fact1 : Fact1( $id1 : id == 1 )
        Fact2( $id2 : id == 2 ) 
        Fact2( id == 2, convertToIdString($id2) == "id2" )  
    then
        System.out.println("***** Action of " + drools.getRule().getName());
        modify($fact1) {
            value = $id2
        }
end

rule "Rule 2"
    when
        Fact1( $id1 : id == 1, value == 2 )
        $fact2 : Fact2( $id2 : id == 2 )
    then
        System.out.println("***** Action of " + drools.getRule().getName());

        modify($fact2) {
            value = $id1
        }
end

Currently:
non-exec-model : property reactive. Not loop
exec-model : class reactive. Infinite loop

=> This is an ambiguous case. $id2 is a bind viable from the previous pattern, so should not trigger property reactive (= exec-model is correct). However, both are the same type so eventually, Fact2.id is the target of property reactive. We would want to meet the exec-model behaviour (plus, compile time warning), but not rushing... we may document the behaviour difference at the moment.

Acceptance Criteria:

• Fix non-exec-model to be class reactive to meet the exec-model behavior.
• Raise a compile time WARN for both non-exec-model and exec-model