Name what each class's instances are a pointer to

[estebanzimanyi]

A binding declares every wrapper in terms of the C type a class stands for, and
the catalog leaves that to the binding to work out — so each carries a map of
its own, and a class the model gains is one the binding cannot type until that
map is edited by hand. classes.<Class>.cType states it instead, over the 21 C
types MEOS's own signatures name.

It is read from those signatures rather than listed. A receiver-role method —
accessor, predicate, conversion, restriction, output — takes the value it is
called on first, so the pointee of that parameter names the type, and the
class's own methods answer for it. A class holding constructors alone says
nothing about itself that way: the concrete <leaf><subtype> classes take the
answer of the subtype they are a product of, TFloatInst reading TInstant
from tinstant_*, and any other class takes its parent's, which is the same C
type by construction. All 105 classes resolve one, and the nine the binding maps
by hand today — Temporal, TInstant, TSequence, TSequenceSet, Set,
Span, SpanSet, TBox, STBox — resolve to what that map says.

Three tests hold the three routes: the receiver, the subtype a constructor-only
class is a product of, and the parent a class with neither answers from.

Stacks on #123.