feat: miner log download via Hypercore P2P streaming

[paragmore]

What are new additions?

Adds a three-step REST flow for downloading miner log files. Since log files
can be large and are generated async by wrk-miner, the client submits a job,
polls until ready, then streams the file directly.

Endpoints:

• POST /auth/miners/:minerId/download-logs — submits a downloadLogs action, returns { jobId, statusUrl, fileUrl }
• GET /auth/miners/:minerId/download-logs/:jobId/status — polls completion; returns pending | ready | failed | expired
• GET /auth/miners/:minerId/download-logs/:jobId/file — streams the binary log
• GET /auth/download-logs/:id — legacy single-step endpoint for existing clients

How is it done ?

Introduces LogDownloader (workers/lib/log-downloader.js), which fetches log bytes directly from the
wrk-miner peer over Hypercore/Hyperswarm P2P — bypassing the HRPC action pipeline entirely. The action
result only carries metadata (coreKey, byteLength, expiresAt); the actual bytes are transferred
lazily as the HTTP client reads the response stream.

A single shared connection handler on net_r0.swarm routes incoming peers to the correct in-flight
download by matching peerInfo.topics against active discoveryKeys. After the stream ends (or errors),
downloaded blocks are cleared from Corestore storage and the session is closed.