Enhancement-606: showmod finds a subcircuit's model by its hierarchical name

Enhancement-606: showmod finds a subcircuit's model by its hierarchical name

Scope: src/frontend/gens.c (dgen_hier_model_match, consulted in dgen_next
beside E-410's instance match), examples/showmodhier_examples/ (new, 8 checks per
solver). ngspice only; built-in and OSDI models alike. Finding N1 of the 2026-09-10
integration hunt
(docs/bug_hunts/2026-09-10_ngspice-osdi-integration-second-hunt.md),
a stock ngspice defect.

Suites: showmodhier_examples 8 of 8 per
solver, both solvers; namelookup, showwidth, instdep, carddefault, multimod unchanged;
full sweep 502 of 502 (one sweep, Enhancements 604–608 folded together).

What was wrong

Subcircuit expansion renames a model card inside x1 to x1:rm (nested:
x1.x2:rm). altermod x1.rm r=3k and altermod x1:rm r=3k both reach it, and
showmod x1.r1 shows it through the instance — but showmod x1.rm and showmod x1:rm
answered "No matching instances or models". The device generator's grammar could
express neither: Enhancement-493's retry compares #x1.rm whole against x1:rm, and
in #x1:rm the ':' is the grammar's subcircuit delimiter, which leaves a device
named rm in subcircuit #x1.

What changed

As Enhancement-410 did for instances, a whole-word match is consulted alongside that
grammar and can only add a match: the query, with or without the model marker #,
equal to the model's name with '.' standing for ':' (either way round, so x1:x2:rm
finds x1.x2:rm). It applies to model queries only — show x1.rm, an instance walk,
finds nothing as before — and needs a '.' or ':' in the query, so a bare showmod rm
keeps its top-level meaning.

Verification

check	result
showmod x1.x2.rm, x1:x2:rm, x1.x2:rm	the model x1.x2:rm, r = 2000
showmod x1.cm, x1:cm	the capacitor model x1:cm
showmod x1.x2.rm : r	the one parameter
altermod x1.x2.rm r=3k, then showmod x1:x2:rm : r	3000; v(out) follows (0.25)
show x1.x2.rm; showmod rm; showmod x1.x2.r1	nothing; nothing; the model through the instance — unchanged
an OSDI model inside a subcircuit, showmod x1.om and x1:om : r	found

Full sweep 502 of 502 on both solvers.

Write-up: enhancements_doc/Enhancement-606.md