[Bug]: query prints per-term guaranteed seeds in alphabetical order, burying the answer file on multi-term queries #4357

Problem

_pick_seeds guarantees one seed per query term (#1445) and appends those seeds in sorted(best_seed_by_term),
which is alphabetical term order. _subgraph_to_text prints seeds first. So on a multi-term query, a weak match on an
early-alphabet term is printed ahead of a specific term's exact match, and the answer file can fall below what an
agent reads first.

A minimal case: the query alpha bravo charlie apple zebra on a graph where all five are exact labels and eight
more labels contain apple. The start line is ['alpha', 'bravo', 'zebra', 'apple', 'charlie'], so the common term's
seed is printed before the rare exact match charlie.

Measured

I measured this on AST-only graphs (graphify update --no-cluster) with graphify's own _score_query, _pick_seeds
and _query_graph_text, on graphify 0.9.86 (v8 a5ba60d). hit@5 = the answer file is among the first 5 files
graphify query prints:

• 114 Django issues from SWE-bench Lite; the answer is the file the official fix changed. Queried by the title plus
  the first 400 characters: hit@5 53. Yet a start node lies in the answer file on 83 of them.
• In those misses the answer's seed usually sits at positions 6-10, after per-term seeds ordered alphabetically.

Proposal

Order the guaranteed seeds by the strength of each term's best match:

• _score_query already ranks each term's winner with (-singleton score, -degree, label length, node id), and the
  singleton score carries the term's IDF weight.
• _score_query would return that order (term_order), and _pick_seeds would iterate it, keeping the sorted
  order when term_order is not given.
• The same seeds are picked, the traversal reaches the same nodes, and the order stays deterministic.

Per question, hit@5 moved +10/-0, +4/-3 and +1/-0 on those three sets. The
evaluation script and data
reproduce every number above.

I have a PR ready, with tests that fail on v8 and pass with the change.