perf(snapshots): Batch image fetches and add timeouts for snapshot download

[NicoHinderling]

Summary

• Batches objectstore image fetches in groups of 200 (was: all at once) to cap memory usage and keep data flowing steadily to the client during streaming ZIP downloads
• Adds 30s per-image fetch timeout to prevent a single slow objectstore read from stalling the entire pipeline
• Adds 90s gateway timeout override for the snapshot download endpoint (matching other large-download endpoints)

Problem

Downloading large snapshots (40K+ images) via GET /api/0/organizations/{org}/preprodartifacts/snapshots/{id}/download/ fails with HTTP/2 stream errors. The endpoint submits all image-fetch futures at once, causing memory spikes and gaps in data flow that trigger connection timeouts.

Fix

• Batched fetching: Process images in groups of FETCH_BATCH_SIZE=200 instead of submitting all futures at once. Each batch completes before the next starts, keeping a steady flow of ZIP data to the client.
• Per-fetch timeout: future.result(timeout=30) prevents indefinite blocking on slow objectstore reads. Timed-out images are skipped (logged as failures), same as other fetch errors.
• Gateway timeout: Added sentry-api-0-organization-preprod-snapshots-download: 90.0 to both sync and async gateway ENDPOINT_TIMEOUT_OVERRIDE dicts.

Test plan

• Existing preprod snapshot API tests pass (37/37)
• Manual test: download a large snapshot (40K images) and verify it completes
• Manual test: download a normal snapshot (400 images) and verify no regression

🤖 Generated with Claude Code

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 55e4647. Configure here.}