edge: add JSON/CSV/Unix-socket output sinks

[armcconnell]

Part 2 of 4 — edge feed parser stack

1. edge: add Top-of-Book parser framework #3522 — edge: add Top-of-Book parser framework ← start here
2. this PR — edge: add JSON/CSV/Unix-socket output sinks
3. edge: add feed runner, manager, HTTP API, and CLI (coming)
4. edge: add integration tests, synthetic publisher, and devnet guide (coming)

Draft until #3522 lands. Reviewers welcome to read for context; I'll mark ready-for-review after rebasing onto main.

Summary

• Introduces the `OutputSink` interface and three implementations: JSON-lines file, CSV file (with header row auto-inferred from the first record's shape), and Unix-socket broadcast with drop-on-slow-consumer semantics so a stuck reader can't backpressure the parser.
• `NewSink` factory keyed on format + path; paths beginning with `unix://` route to the socket sink, others to file sinks.
• CSV sink pivots the parser's flat `Fields` map onto a stable column set so heterogeneous record types (quotes, trades, instrument definitions, heartbeats) line up in one sheet.
• The parser still knows nothing about sinks — these are pure consumers of `[]Record` produced by the parser from edge: add Top-of-Book parser framework #3522.

Testing Verification

• Unit tests cover each sink: CSV header inference + column-pivot, JSON line format, socket broadcast to multiple connected clients, and slow-consumer drop behavior.