Fix how empty function pointer cast tables are handled

[mdboom]

This fixes what I think is a fairly straightforward bug when EMULATE_FUNCTION_POINTER_CASTS is turned on.

When compiling Python as here, some of the tables end up having 0 length. In that case, they look like [] and when fixup_metadata_tables tries to make all the tables the same length, it sets these empty ones to [,0,0] rather than [0,0] as it should. This triggers the assertion:

          assert i not in function_pointer_targets

later on in make_function_tables_defs, since many of the tables have the same empty string in the first entry in the table.

This patch seems to get me to a working state, but I don't know if having empty tables in the first place is the sign of some deeper issue.

[kripken]

Looks good, thanks! And the CI failure looks like a random bot problem we can ignore.

Please just add yourself to AUTHORS before we merge.

Empty tables is odd, but not necessarily a bug. It can be caused by doing a function pointer call but not having a function implemented of that type - perhaps such a function is expected to arrive from a dynamically linked module.

[mdboom]

Thanks. I've added my name to AUTHORS.

[kripken]

Thanks again!