Builds the text a cross-encoder reranker scores for a candidate.

Faithful port of Hindsight's CrossEncoderReranker.rerank doc_text assembly (engine/search/reranking.py lines 188-210): layer context and date onto the content so the reranker has temporal + contextual signal, not bare text.

text                                   → "walked the dog"
context: "diary"                       → "diary: walked the dog"
occurred_start: 2022-06-05             → "[Date: June 05, 2022 (2022-06-05)] diary: walked the dog"

Each layer is added only when present; the date (from occurred_start) wraps the whole thing including context, exactly as the reference does.