Stream object bytes from R2 using the S3 API (not HTTP fetch to a presigned URL).
Each call performs its own GetObject and returns an independent Web ReadableStream.
That lets multiple platforms read the same object in parallel without sharing a
single fetch() Response body (which can trigger "body disturbed or locked"), and
without buffering the entire object in memory (important for multi‑GB files).
Stream object bytes from R2 using the S3 API (not HTTP fetch to a presigned URL).
Each call performs its own GetObject and returns an independent Web ReadableStream. That lets multiple platforms read the same object in parallel without sharing a single fetch() Response body (which can trigger "body disturbed or locked"), and without buffering the entire object in memory (important for multi‑GB files).