The current MicroClaw release artifact is substantially larger than necessary for common deployments because channel integrations are statically compiled into the default binary even when they are not used.

Observed impact:

• The default release binary pulls in Telegram, Discord, Matrix, Web, and other channel-specific dependency trees together.
• Heavy channel stacks such as Matrix contribute a large amount of code and transitive dependencies.
• Users who only need a subset of channels still pay the full artifact size cost.

Follow-up work proposed in this issue:

• Introduce Cargo features for channel-specific integrations.
• Make channel-specific dependencies optional.
• Gate channel modules and runtime startup paths behind those features.
• Define smaller default builds and keep a separate full build for all channels.

This issue intentionally excludes the immediate release-profile size optimization that is being handled separately in an accompanying pull request.