ENG-175: move webhook parsing + MCP discovery to worker; remove backend docker

[betterclever]

Addresses Linear ENG-175.

Changes:

• Webhook parsing runs via Temporal in the worker (backend no longer runs parsing/docker directly).
• Removed all backend direct Docker spawning for MCP discovery and made controller leaner.
• MCP library UI cleanup: remove duplicate group import button, simplify group import UX.
• Fix: MCP group tool counts show immediately after import (no refresh needed).

Testing:

• bun run typecheck
• bun run test
• RUN_E2E=true bun test e2e-tests/webhooks.test.ts

Linear: https://linear.app/shipsec-ai/issue/ENG-175/webhook-parsing-test-fails-docker-api-connection-error-no-such-file-or

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ca801103c7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Disable automatic retries for webhook parsing activity

When the user parsing script throws (syntax/runtime error), webhookParsingWorkflow will keep retrying executeWebhookParsingScriptActivity with Temporal's default retry policy because only startToCloseTimeout is set here and no retry policy limits attempts. In this flow the backend blocks on temporalService.getWorkflowResult(...) in WebhooksService.executeParsingScript, so a permanently failing script can leave webhook deliveries stuck in processing and the request hanging until infrastructure timeouts instead of failing fast. Please set retry.maximumAttempts (typically 1) or a bounded scheduleToCloseTimeout for this activity.

Useful? React with 👍 / 👎.

[betterclever]

CI lint failures were prettier/prettier formatting issues. Ran eslint --fix (backend + worker) and pushed.

Also addressed potential infinite retry/hanging webhook parsing: webhookParsingWorkflow now bounds activity retries (maximumAttempts=1) and uses a bounded scheduleToCloseTimeout/startToCloseTimeout derived from input.timeoutSeconds (default 120s, capped at 10m).