fix: return 500 on internal failures

[jranson]

Description

This fixes the remaining cases outlined in #1018 - when a 200 response is received from a backend but trickster internally fails to handle the request (e.g., unmarshal failure). Other cases where a 200 was being returned instead of 5xx are already addressed. With this change, instead of proxying the 200 response to the caller on internal failures, the response code is now changed to 500 Internal Server Error before responding. When the internal failure is on one or more shards of a TSM operation, the TSM handles the failure(s) the same as if the failure came from the remote origin. Related tests are updated with the new expectation.

Type of Change

• • Bug fix
• • New feature
• • Optimization
• • Test coverage
• • Documentation
• • Infrastructure

AI Disclosure

• • This contribution DOES NOT include AI-generated changes
• • This contribution DOES include AI-generated changes, and I have reviewed the relevant contributing guidelines.

[houyuwushang]

One small question: should dpcProxyErrorStatusCode also normalize 3xx statuses? fetchExtents treats any resp.StatusCode != http.StatusOK as a failed extent, so an all-3xx DPC failure would still be recorded as proxy-error but respond with that 3xx instead of a 5xx. If preserving upstream redirects/statuses is intentional here, this is fine; I just wanted to check because #1018 frames proxy-error responses as needing a 5xx.

[jranson]

Currently none of the supported backends that use DPC (prometheus, influxdb, clickhouse) return 3xx responses to a Trickster request (no redirects, and Trickster does not send If-Modified-Since headers to time series backends), so this should not be an issue.