Fix MCP status drift after auth/restart/reload

[adityakulraj]

Summary

• normalize MCP status mapping in runtime and workspace hooks
• always apply MCP status responses (including empty server lists) so removed/disconnected servers clear correctly
• preserve MCP error details when mapping statuses
• avoid falling back to draft statuses for live sessions, preventing stale status leakage

Validation

• confirmed no remaining guarded status updates (result.servers?.length / statusResult.servers?.length) in the updated hooks
• could not run project build locally because pnpm is unavailable in this environment and corepack pnpm fails due to TLS certificate verification

Closes #38