Add opt-in early skips for draft and non-code-only pull requests

[asdf8675309]

Summary

Repeated CodeQL runs of about five minutes during rapid draft-PR iteration occupy runners before the author is ready for review. On limited capacity runners, those runs can also delay merge-queue validation. This change adds opt-in early exits for confirmed draft PRs and conservatively identified non-code-only PR changes.

Couldn't find a way to not have the Code Quality scan run on draft PR's and PR's that didn't change any of the file types that were scanned. This adds skip-if-draft and skip-if-no-language-changes to init, both defaulting to false.

• Check confirmed draft state before GitHub API setup, repository-property/configuration lookups, and starting/completed status reports. This avoids telemetry requests and their retries as well as CodeQL CLI download and database initialization.
• For the change gate, require one explicit built-in language and immutable PR event base/head SHAs. Include both sides of renames and decline to skip when the comparison may be truncated (300 or more files), fails, or contains unknown paths. Only recognized non-code paths qualify; source, workflow, build/dependency configuration, and shell-script changes retain analysis.
• Preserve complete language analysis whenever a scan runs. This does not extract or query only changed files.
• Export analysis-skipped and analysis-skip-reason; make autobuild, analyze, and their post actions handle intentional skips successfully. Custom build steps can use the output to guard their own work.
• Add usage documentation, a change note, regression tests, and regenerated compiled artifacts.

A workflow enabling draft skips must include ready_for_review in its PR event types. The path gate is a conservative heuristic: repositories that generate code from documentation or other allowed non-code inputs should leave it disabled.

Managed Code Quality integration and review questions

GitHub-managed Code Quality uses a generated dynamic workflow. This PR supplies an action-side mechanism; it does not modify that generator or the product's repository settings. The proposed CODE_SCANNING_IS_DRAFT environment signal allows the generator to supply draft state. Managed adoption would also need the appropriate opt-in inputs, immutable PR commit metadata for the change gate, and ready-for-review scheduling.

Feedback is particularly welcome on:

1. Whether these opt-in controls belong in the action, and whether the input names and proposed managed draft signal fit existing integration conventions.
2. Whether confirmed draft skips should deliberately omit start/completion status telemetry. The job remains successful and the logs/outputs state explicitly that analysis was skipped.
3. Whether the initial non-code allowlist is appropriately conservative, or a narrower policy should ship first.

DRAFT risk assessment from Codex --

Proposed classification for maintainer review:

• High risk: The new inputs are disabled by default, but enabling them intentionally omits analysis. A false skip is possible when recognized non-code paths are inputs to generated code. The changed-file gate and managed integration have limited live validation, and confirmed draft skips omit status telemetry. The README documents the path-policy limitation and required ready-for-review trigger.

Which use cases does this change impact?

Workflow types:

• Advanced setup - Impacts users who have custom CodeQL workflows and enable the new inputs.
• Managed - Provides a proposed integration point for dynamic workflows. Current generators are not changed by this PR; adoption requires GitHub-side integration.