feat: Added compact command

[michaelbarnes]

This PR introduces the compact command to the CLI allowing developers to run compact jobs on demand via the CLI.

Usage

powersync compact
# or with explicit IDs (no link required):
powersync compact --instance-id <id> --project-id <project> --org-id <org>

Implementation

The command is modeled directly on powersync destroy:

• Extends CloudInstanceCommand, so it reuses the existing instance/project/org resolution from cli.yaml, flags, or env vars.
• Calls client.compact({ org_id, app_id, id }).
• If the response includes an operation_id, polls via waitForOperationStatusChange with a 30-minute timeout (longer than destroy's 10 min, since compaction can run long on large buckets). If the response has no operation_id, treats the call as immediately complete.
• Reports Instance compacted successfully, on success, or directs users to powersync status for diagnostics on failure.

This is a Cloud-only command.

Tests

Added cli/test/commands/compact.test.ts covering:

• success path (with and without operation_id in the response)
• polling reports failed status
• network/client error
• missing-link (no cli.yaml) and self-hosted project rejection

All 127 tests in the CLI suite pass.

Incidental cleanup

Fixed two pre-existing lint errors in cli/test/commands/migrate.test.ts that were blocking the pre-commit hook:

• Wrapped the top-level test() in a describe('migrate', () => { it(...) }) to clear
  mocha/no-global-tests.
• Changed 'utf-8' → 'utf8' to clear unicorn/text-encoding-identifier-case.

AI Disclaimer

Claude Opus 4.7 assisted in the development of this feature.

[stevensJourney]

Happy with these changes :)

[stevensJourney]

Compact operations can generally take quite some time. For that reason, I can see many more potential occurrences for this command to timeout - compared to the deploy command which is a less likely to timeout in a standard timeframe. I think it could be nice to potentially make this timeout configurable with an optional input flag.