fix: accessor map keys must match embedded callable signatures

[rahlk]

Problem

Getter/setter callables are stored under access-qualified map keys, but the embedded TSCallable.signature is left unqualified. This violates the otherwise central identity invariant that a callable's map key byte-matches its signature and the IDs used by call edges.

It also permits a getter and setter for the same property to advertise the same embedded signature even though their map keys differ.

Reproduction

OWASP Juice Shop 6244c59, current v1 JSON output:

• 11 getter records have a map key ending in #get;
• all 11 embedded signature values omit #get.

Example shape:

{
  "SomeClass.value#get": {
    "kind": "getter",
    "accessor_kind": "getter",
    "signature": "SomeClass.value"
  }
}

The implementation makes the mismatch explicit in src/syntactic_analysis/builders.ts:

const r = buildCallable(g, g, "getter", root);
if (r) methods[`${r.sig}#get`] = r.callable;

The setter branch has the same form with #set.

Expected behavior

A callable has one canonical ID. Its container key, embedded signature, callsite resolution, call-graph endpoint, cache identity, and Neo4j signature must agree exactly.

Acceptance criteria

• Pick one canonical accessor identity strategy:
  • include #get / #set in both key and embedded signature; or
  • represent accessors without suffixes using a collision-safe alternative.
• Ensure getter and setter declarations for the same property remain distinct.
• Update caller/callee canonicalization and Neo4j projection to use the same ID.
• Add a regression fixture containing both get value() and set value(...) and assert:
  • map key equals callable.signature for both;
  • IDs are distinct;
  • any accessor-related graph references target the corresponding declared ID;
  • eager and cached output use identical IDs.