feat(parser): ingest transcripts and dedupe by session id

[hora7ce]

Description

This PR adds transcript ingestion support to VS Code parsing and safely deduplicates transcript/chat session overlap by normalized sessionId.

Changes included:

• Added transcript parsing support and request construction from transcript events
• Added tool extraction and dedup helpers for transcript tool-call metadata
• Added merge logic to dedupe transcript sessions with existing chat sessions
• Scoped dedup indexing to workspace-local parsing context in sync and async paths
• Added tests for single-turn/multi-turn transcript parsing, invalid/empty transcript handling, tool dedup, transcript-chat dedup, and transcript-only fallback ingestion

Related Issues

Relates to #87
Supersedes prior context in #64

Checklist

• npm run check passes (typecheck + lint + spellcheck + knip + tests)
• Changes are covered by tests (if applicable)
• Documentation updated (if applicable)

[hora7ce]

Reviewer context against issue requirements (#87, #64):

This change is intentionally aligned to the investigation direction in #87 and the acceptance criteria in #64.

Requirement coverage:

• Dual-source ingestion without duplicate sessions:
  • Transcript sessions are merged by normalized sessionId with existing chatSessions data.
  • Hard rule from Investigation: dual-source session ingestion (chatSessions/ vs GitHub.copilot-chat/transcripts/) — dedup strategy needed before adopting transcript parser #87 is preserved: one Session object per sessionId across sources.
• Canonical source + enrichment/recovery behavior:
  • Existing chatSessions parsing remains primary.
  • transcripts/*.jsonl are ingested as secondary input and merged when the same session exists, or added as fallback when transcript-only.
• Transcript event-stream parsing:
  • session.start / user.message / assistant.message and tool execution events are mapped into SessionData requests.
  • Tool names are deduplicated per turn.
• Graceful handling of bad input:
  • Empty/malformed transcript files return null and are skipped.

Test coverage added for the above:

• parseTranscriptFile:
  • single-turn parse
  • multi-turn parse
  • empty/corrupt input => null
  • per-turn tool dedup
• processWorkspaceEntry dedup flow:
  • chat + transcript with same sessionId emits one merged session
  • transcript-only session ingestion works

Validation status:

• npm run check passed locally on this branch (typecheck + lint + spellcheck + knip + tests)

Notes for reviewers:

• This directly addresses the duplicate-session regression concern called out in Investigation: dual-source session ingestion (chatSessions/ vs GitHub.copilot-chat/transcripts/) — dedup strategy needed before adopting transcript parser #87 / PR feat: parse GitHub.copilot-chat/transcripts/*.jsonl event-stream format #70.
• Engineering plan/review/test docs were prepared during implementation but intentionally left out of this code-only PR scope.