feat(scrolls): expose download progress across catalog

[IOHelpMe]

Summary

• instrument every published SteamCMD-based LGSM and Rust Scroll through the shared druid progress steamcmd adapter
• route measurable Minecraft, Cuberite, Oxide, and Hytale HTTP payloads through the central druid progress download producer
• give Forge and non-measurable Hytale work stable, truthful opaque installation phases instead of fake percentages
• pin progress-capable procedure images to v0.1.258 instead of mutable stable tags
• retarget embedded procedure images when publishing local or preview artifacts
• make plain-HTTP local seeding use druid:local and druid:local-steamcmd
• enforce catalog-wide classification so a newly published Scroll cannot silently bypass the progress contract

Dependency and merge order

This consumes the producers added by highcard-dev/druid-cli#97.

1. release druid-cli#97 as v0.1.258
2. merge this PR
3. merge highcard-dev/monorepo#318

Validation

• go test -count=1 ./...
• bash -n scripts/push.sh
• fake publication verifies embedded images for explicit preview targets and the plain-HTTP local seed path
• catalog regression validates every released Scroll family
• make -n k3d-build-pull-image confirms both required local images
• git diff --check

End-to-end evidence

• Cuberite exercised the real HTTP producer and reached Active
• Project Zomboid exercised the real SteamCMD producer and completed installation
• a realistic sparse SteamCMD trace verified speed, gated ETA, stall handling, and recovery without another multi-gigabyte game download

Backend and SPA PRs:

• feat(runtime): expose precise SnapshotProgress during setup druid-cli#97
• https://github.com/highcard-dev/monorepo/pull/318

[druid-infra]

Error: This repo is not allowlisted for Atlantis.

[IOHelpMe]

GPT-5.6 implementation is in 86ad46d2643953ed5907de35fc642b06718fdbd6.

Only the ARK update and auto-install procedures are instrumented. The wrapper preserves the existing LGSM/SteamCMD output and process result while emitting validated download/verify/install measurements consumed by highcard-dev/druid-cli#97.

Fresh verification:

• go test ./... — passed
• bash scripts/validate_all_scrolls.sh — the complete catalog passed
• git diff --check — passed
• fresh local ARK deployment — approximately 9m15s installation, 249 progress updates, then successful launch to Active

Merge order: highcard-dev/druid-cli#97 → this PR → highcard-dev/monorepo#318.

[IOHelpMe]

@MarcStdt this is ready for review. GitHub did not allow IOHelpMe to create a formal review request on this cross-fork PR; it depends on highcard-dev/druid-cli#97 and should merge before highcard-dev/monorepo#318.

[IOHelpMe]

Follow-up after senior review:

• expanded progress coverage from ARK to every published SteamCMD/HTTP-capable Scroll
• added truthful opaque phases where a total is not measurable
• removed mutable progress runtime tags, pinned the v0.1.258 contract, and fixed local/preview artifact retargeting
• added catalog-wide and fake-publication regression coverage; full Scrolls Go suite passes

Merge gate: druid-cli#97 must be released as v0.1.258 before this PR is merged/published.

Latest fix commit: 90ef3b21.

[IOHelpMe]

Closing because HIG-24 was deliberately reduced to the existing SnapshotProgress signal. The clean rewrite requires no Scroll catalog instrumentation, SteamCMD parsing, HTTP download wrappers, or Scroll-specific changes; the relevant implementation now lives only in druid-cli#97 and monorepo#318.