Replace all uses of type_eq by base_type_eq

[tautschnig]

It holds that type_eq(t1, t2, ns) implies base_type_eq(t1, t2, ns), but the converse isn't necessarily true (if expansion of symbol types isn't consistent between t1 and t2). This distinction, however, is subtle and only comes with a some (not measured) performance benefit. Maintaining two functions doing almost the same is painful and confusing.

In the medium term, neither type_eq nor base_type_eq should be necessary as tag types ought to enable use of operator== as the one and only equality comparison over types.

[tautschnig]

type_eq has been removed in #4304.