Support Large Files

[philomath213]

Summary

1. File uploads previously read the entire file into memory with f.read() before POSTing, causing socket.send() to fail on large files.
2. Now passes the open file handle directly to requests, which streams the body in chunks.
3. HMAC-SHA256 signing is preserved by hashing the file incrementally (64 KB chunks) before seeking back to 0.

Test plan

• pytest passes
• Upload a large ZIP file (bhcli upload large.zip) and confirm it completes without a socket/buffer error

[exploide]

Thanks for your improvement. Will merge soon.

Out of interest, how large was your file that caused a problem?

And I'm thinking about the hashing in 64k chunks. The hashing function can easily process much larger chunks and I'm wondering whether thousands of read-64k iterations would cause a certain IO overhead. But probably not that important.