SSH: Optimize pack data handling for large repositories

[dcoric]

Description:

Current pack data capture approach buffers all data in memory, which could cause issues with large repositories.

Current Approach:

const packDataChunks: Buffer[] = [];
// ... accumulate all chunks in memory
const packData = Buffer.concat(packDataChunks);

Potential Issues:

• Memory exhaustion with large pushes (>500MB)
• No progress reporting for slow connections
• All-or-nothing approach (no partial processing)

Proposed Optimizations:

1. Streaming to temporary file:

   • Write chunks to temporary file instead of memory
   • Process from file after stream completes
   • Automatic cleanup on completion/error

2. Size limits:

   • Configure maximum pack file size
   • Graceful degradation for oversized packs
   • Clear error messages when limits exceeded

3. Progress reporting:

   • Track bytes received vs expected
   • Timeout for stalled transfers
   • User-facing progress indicators

Affected Files:

• src/proxy/ssh/server.ts (stream handling)
• src/config/ (size limit configuration)