…tension

Adds Amazon Redshift as a queryable destination in the Workflow Insight VS Code
extension, alongside the existing DynamoDB / Aurora / S3-Athena / CloudWatch
Logs destinations. Redshift is modeled on the Aurora path (SQL over the same
record_json + scalar-columns table) with the Redshift-specific differences
handled:

- New src/redshift.ts: runRedshiftQuery + fetchRedshiftRecord over the Redshift
  Data API. Unlike the RDS Data API, the Redshift Data API is asynchronous, so
  it executes then polls DescribeStatement to completion and pages through
  GetStatementResult, normalizing to the shared result shape.
- schema.ts: new "redshift" DestinationType and a dedicated NL->SQL prompt
  (record schema, dialect, few-shots) covering SUPER/PartiQL navigation. JSON
  keys are camelCase, so the prompt requires double-quoted SUPER paths
  (record_json."functionName", o."durationMs") and array UNNEST via a joined
  table alias.
- config.ts / package.json settings: redshift workgroup/cluster, database,
  schema, table, dbUser, and secretArn (Serverless or provisioned). Wired
  through extension.ts (query + record-fetch routing) and destinationTest.ts
  (SELECT 1 connectivity probe). Settings webview updated (SettingsModal, types,
  starter query). Adds @aws-sdk/client-redshift-data.

Example CDK stack + exporter fixes (validated end-to-end against a live Redshift
Serverless workgroup):

- RedshiftExporter: rewrite the upsert MERGE to project the row in the USING
  subquery and join on src.execution_arn. The previous form joined on a bound
  parameter, which Redshift plans as a NestedLoop join and rejects with
  "NestedLoop join is not supported in MERGE" — so every export silently failed.
- redshift-create-table handler: run multiple statements (splits on ';') so the
  provisioning SQL can CREATE TABLE and GRANT in one custom resource.
- stack.ts: GRANT ALL on the table TO PUBLIC (the table is owned by the
  create-table role's DB user; the exporter and query identities are different
  users and were otherwise denied), set enable_case_sensitive_identifier=true on
  the workgroup so camelCase SUPER keys are reachable, and pass the Redshift
  env vars to the example function so it emits records via RedshiftExporter.

Tests: unit tests for runRedshiftQuery/fetchRedshiftRecord (polling,
pagination, failure, SUPER parse) and extended config/destinationTest coverage.