[feaLib] _upgrade_mixed_subst_statements does not account for all possible lookup breaks #3846
Description
Another little issue with #3805.
Consider the following FEA:
feature derp {
sub x by y;
lookupflag 4;
sub f i by one;
} derp;because the logic in up (
fonttools/Lib/fontTools/feaLib/ast.py
Line 340 in b21e538
sub x by y rule into a ligature lookup, even though the subsequent lookupflag statement means it is not in the same lookup as the subsequent ligature sub rule.
The same miscompilation can also occur, for instance, with FEA like:
lookup ther {
sub l by j;
} ther ;
feature clig {
sub a by b;
sub e l l' lookup lookup_1;
sub e l f by z;
} clig;where the contextual rule in the middle forces a new lookup, and we end up giving the first rule the wrong lookup type.
(ultimately this is another one where the question is, 'do we care'?)