cmpltv: don't raise from print-object on a half-built creator

[dg1sbg]

Two small robustness changes in src/lisp/kernel/cmp/cmpltv.lisp. Neither fixes a bug I can currently demonstrate; both were found while chasing a failure whose real cause the first one was actively hiding.

print-object can raise and swallow the condition it was printing

print-object for vcreator already guards %prototype with slot-boundp, but reads (index object) unguarded three lines later — and the creator method reads it unguarded too. A creator whose index has not yet been assigned therefore signals an unbound-slot error from inside the printer, which replaces whatever condition was actually being reported and splices the new message into the middle of the old one:

#<CMPLTV::BASE-STRING-CREATOR [no prototype]The slot CMPLTV::%INDEX is
                                            unbound in an instance of
                                            CMPLTV::BASE-STRING-CREATOR.

That is what the failure I was investigating looked like, and it cost real time — the visible error is entirely about the printer, while the original condition is gone. This guards both reads the way %prototype already is, so such an object prints as [no index] and the real condition survives.

%index type contradicts its own initform

The slot is declared :type (integer 0) with :initform nil, which the type excludes. Upstream Maclina has :type (or null (integer 0)) for the same slot; this looks like drift. Nothing enforces it today — I checked, make-instance stores nil there without complaint — so this is tidiness rather than a fix, but the declaration should not contradict the initform.

Verification

x86-64 Linux, LLVM 18, --build-mode=bytecode: builds clean, regression suite 1963 successes / TEST_EXIT=0, identical to baseline.

[dg1sbg]

Note on the two CI reds — both are the same known flake, and neither relates to this change.

clasp/ubuntu-latest/native and cando/ubuntu-latest/native each failed with WEAK-KEY-AND-VALUE-WEAKNESS as the sole failure, Successes: 1962, and the normal 2 aborted compilation units. That is #1814, where I have posted frequency data — it accounted for seven Linux job failures across five unrelated PRs today.

This PR touches only print-object methods and one slot type declaration in cmpltv.lisp; it has no plausible connection to weak hash tables or the GC.

clasp/ubuntu-latest/bytecode passed, and I separately verified this change on x86-64 Linux: clean build, regression suite 1963 successes / TEST_EXIT=0, identical to baseline.