[snapshot-runtime] Harden Snapshot Download Pipeline for Fork Deploy Reliability

[kangeunchan]

Enhancement Description

• One-line description (release-note style): Harden snapshot download pipeline with resume, parallel range fetch, and clearer progress/failure semantics for Cosmos fork deploy reliability.
• Scope type: Snapshot download and extraction reliability hardening

Background

Snapshot download became a critical bottleneck/failure point for deploy --fork on Cosmos mainnet/testnet.

This issue tracks:

• Download-path reliability improvements that were implemented beyond baseline plugin completion scope

Scope

• Add parallel range download with single-stream fallback
  • File: internal/infrastructure/snapshot/download.go
  • Includes range capability probe, fallback behavior, and safe finalization
• Add resume support and temporary file continuation behavior
  • File: internal/infrastructure/snapshot/download.go
  • Includes .tmp continuation and server response handling
• Improve progress reporting and timeout handling for long-running snapshot downloads
  • Files: internal/infrastructure/snapshot/download.go, internal/application/devnet/progress_bar.go
• Refactor snapshot download flow for clearer execution contract and tests
  • Files: internal/infrastructure/snapshot/download.go, internal/infrastructure/snapshot/download_test.go
• Expose snapshot timeout on deploy path
  • Files: cmd/devnet-builder/commands/manage/deploy.go, internal/application/dto/devnet_dto.go, internal/application/devnet/provision.go, internal/application/devnet/run.go

Non-Goals

• Replacing external snapshot providers
• Changing Cosmos plugin business logic (genesis mutation/RPC semantics)
• Introducing a generalized multi-chain snapshot framework

Risks and Open Questions

• Provider-side per-connection throttling may still dominate throughput
• Resume behavior differs across providers/CDNs that do not honor Range consistently
• Locking/concurrency policy for same cache key in concurrent deploy invocations may need stricter guarantees

Validation Plan

Unit and Integration Checks

• ASDF_GOLANG_VERSION=1.24.0 go test ./internal/infrastructure/snapshot/...
• ASDF_GOLANG_VERSION=1.24.0 go test ./internal/application/devnet/...
• Add explicit tests for concurrent same-cache-key download contention

End-to-End Checks

• devnet-builder deploy --blockchain cosmos --network mainnet --fork --snapshot-timeout 2h
• devnet-builder deploy --blockchain cosmos --network testnet --fork --snapshot-timeout 2h
• Validate resume behavior by interrupting download and rerunning deploy

Evidence Required in Issue Updates

• Download throughput/progress evidence for mainnet snapshot
• Resume continuation evidence (.tmp reused) after interruption
• Fallback path evidence when range/parallel is not available

Acceptance Criteria

• Snapshot download succeeds reliably under long-running network conditions
• Interrupted downloads can resume without full restart when provider supports Range
• Parallel mode improves time-to-download where provider/network allows
• Fallback mode remains correct and deterministic when parallel is unavailable

Deliverables

• Pushed snapshot download pipeline improvements branch for review
• Pushed unit tests covering range/resume/fallback behaviors for review
• Follow-up notes documenting provider-specific throughput/fallback observations