fix(resolution): resolve Python module members through an aliased from-import (#1626)

[maxmilian]

Fixes the from … import … as … half of #1626.

resolvePythonModuleMember rebuilds the submodule's dotted path by joining the import source with the local name. Under from pkg import mod as alias that produces pkg.alias — a module that doesn't exist — so the file lookup finds nothing, the member never resolves, and the call lands in unresolved_refs with status='failed'. codegraph_callers then reports the target as having fewer callers than it does, which is exactly the wrong "is this dead code?" answer that #578 produced for the unaliased form.

Joining with the exported name fixes it. For an unaliased import the two names are identical, so nothing changes there; the namespace form (exportedName === '*') keeps binding at the local name, which is what it already did.

Scope: the report has two halves and only one reproduces

The issue also lists import build_api_contract as builder → builder.source_fingerprint() as producing no edge. That form resolves on current main. It's a namespace import that binds at source — the real module name — so it never goes through the local-name join this PR fixes.

I checked rather than assumed: with the resolver change reverted, a probe on the plain-aliased call site in the test fixture still reports one calls edge. So if @JoeyNPP is still seeing that case fail on a real project, there's a second cause and it isn't this one — worth keeping #1626 open on that half, or splitting it out.

Tests

One case added beside the existing #578 test in resolution.test.ts (same integration shape — real CodeGraph.init with index: true), asserting both halves:

• from pkg import module as mod_alias → mod_alias.func() resolves to pkg/module.py — fails before the change (expected [] to have a length of 1)
• import top_level as tl → tl.top_func() resolves — green before and after, pinned so fixing one half can't break the other

Full suite: 178 files / 3053 tests passing (3052 on main plus this one). tsc --noEmit clean.

Reported by @JoeyNPP.