There are two changes here. One is to pass tracking=True to our parser. This makes it properly track positions
of all productions, not just of lexer tokens. The second is to properly count up our newlines in the lex data,
since the lexer seems to report the line number of the start of the lex data, which is always 1 in our case.
The big block in getRetvalDeclarationForType is just direct cut/paste from CGCallGenerator plus the addition of the sequence case. The IDL parser changes were OKed by khuey; they're needed so that we don't have to worry about the ordering of sequence with conversions for strings and primitives.
There are two parts to this change:
1) Sequences and arrays are now distinguishable from all non-callback interfaces.
2) To handle that change, overload resolution first tries to convert to the exact
interface, and only if that fails tries to convert things with indexed properties
to sequences or arrays.
