DiscussionService flat active-tier and lazy ordinal archive refactor (#11187 AC5)

Context

Part of Epic #11187 Phase 1 execution. We are converging on the Option E''+S architecture for GitHub workflow data archiving: discussions need to be collapsed to a flat structure in the active tier, and use the new lazy ordinal chunking in the archive tier.

The Problem

The current DiscussionService and DiscussionSyncer write to sparse XXxx directories because the type-density for discussions is very low (~68 total items). A 100-ID chunking model produces 1-5 item folders, which is a structural anti-pattern. Moreover, no formal discussion-archive/ layer currently exists.

The Architectural Reality

• ai/services/github-workflow/DiscussionService.mjs handles file operations for Discussions.
• The active tier must abandon sparse XXxx chunking and write directly to a flat discussions/ directory.
• The archive tier must use the new archivePath() helper (AC2) pointing to the unified archiveRoot specified in the config (AC1). Given the current density, the helper will inherently yield a flat folder (archive/discussions/vN.M.K/), but the hook enables future scaling.
• LocalFileService.mjs currently lacks native support for discussion lookups (unlike getIssueById), which is now required to resolve the active vs archive tier splits.

The Fix

Refactor DiscussionService to collapse active discussion paths to a flat directory. Wire the archive-write path to use the archivePath() helper, establishing the formal archive destination for discussions. Add getDiscussionById to LocalFileService.mjs to support querying the new dual-tier structure.

Contract Ledger Matrix

Acceptance Criteria

• Refactor active-tier discussion writes to use a flat directory structure, eliminating the sparse XXxx chunking.
• Implement the archive-tier write path for discussions using the new archivePath() helper.
• Implement LocalFileService.getDiscussionById to correctly resolve flat active discussions and fallback to archive resolution.

Out of Scope

• Actually migrating existing XXxx discussion files to flat structure (Phase 2 AC6).
• Actually creating the historical archive for discussions (Phase 3 AC9).
• Modifying IssueService (AC3) or PullRequestService (AC4).

Avoided Traps

• Using ordinal chunking for the active tier. Even though we are flattening it, if it ever scales, the active tier must use deterministic ID-range chunking, never insertion-order ordinals. For now, flat is correct for its density.

Origin Session ID

Origin Session ID: 57502eb2-7f7b-4b9b-a849-49f016b08c95 Retrieval Hint: "Epic 11187 Phase 1 DiscussionService Refactor"