fix(schema): decorator qualified_name is the import-table resolution or absent; name is the written spelling #151

Problem

TSDecorator.qualified_name is documented as "checker-resolved FQN when available". It is not
resolved by anything. It is ts-morph's Decorator.getFullName() — the written expression text —
and the builder never consults the checker (src/syntactic_analysis/builders.ts, decoratorsOf).

Measured with a decorator imported as import { Get as HttpGet } from "./decorators", where
resolution is trivial: the field emits HttpGet, the alias as typed. So the field is a second copy
of name with dots, and every consumer that trusted the comment — including #143 — was misled.

Consequence in #143 as shipped: :TSDecorator merges on qualified_name || name, so a project's
local @Get and NestJS's @Get collapse into one node. The "two spellings, one decorator"
behaviour that PR describes works in the opposite direction from python's.

Found while propagating python 1.4.1's entrypoint changes; this is unit 0 of
docs/design/specs/entrypoint-detection.md
(#150), and blocks the decorator matcher there.

Scope boundary

The two decorator identity fields only. No entrypoint detection, no rules file, no changes to
argument capture.

Goals

• qualified_name is the import-table resolution of the written spelling, or absent when
  the decorator's head is not an imported binding — python's rule ("Jedi, else the import table,
  else None") minus the Jedi step TypeScript does not have
• name is the decorator as written (http.route, not route), so an unresolved dotted
  decorator keeps its spelling in the output — python's name is the written spelling for the
  same reason
• The resolver is reusable: the entrypoint pass (units 2-3) needs the same import table for base
  classes and the unresolved counter
• Schema comments say what the fields actually hold

Caveats and known risks

This changes emitted values, so it is a fix with a visible effect rather than an invisible one:

Neo4j :TSDecorator merge key (qualified_name || name) therefore becomes package-qualified for
imported decorators and stays the written spelling for local ones — which is what #143 meant.

Not a breaking change: the field was documented as resolved and never was. But anything that
matched on the old bare qualified_name sees different values.

Definition of done

• The four rows above are asserted end-to-end, including the alias and namespace cases
• A same-file decorator emits no qualified_name
• bun test green; the feat(neo4j): project decorators, matching python #143 test still passes (its decorators are same-file, so the merge key is
  unchanged there)
• Container tests run against live Neo4j