[jcjones]

Recent DAP drafts only require "a successful status code", so collection job creation and polling no longer insist on 200 OK, which we need to reflect for interoperability.

Location	Request	Before	After
collector/src/lib.rs:713 (send_collection_request)	Collector → Leader: create collection job (POST)	Exactly 200 OK	Changed: any 2xx
collector/src/lib.rs:763 (poll_once)	Collector → Leader: poll collection job (GET)	Exactly 200 OK	Changed: any 2xx except 204, which Janus leaders send for deleted jobs
collector/src/lib.rs:956 (delete_collection_job)	Collector → Leader: delete collection job (DELETE)	Any 2xx	Unchanged
client/src/lib.rs:744	Client → Leader: upload reports	Any 2xx	Unchanged
client/src/lib.rs:854	Client → OHTTP relay: upload over OHTTP	Any 2xx	Unchanged
core/src/http/cached_resource.rs:140	Client → Aggregators: fetch HPKE configs	Any 2xx	Unchanged
core/src/retries.rs:358 (retry_http_request), used by send_request_to_helper (aggregator/src/aggregator.rs:4206)	Leader → Helper: aggregation jobs and aggregate shares	Only 4xx/5xx are errors	Unchanged

The leader answers a poll on a deleted collection job with 204 No Content and an empty body. If the collector treated that as success, it would read the empty body as "not ready" and poll forever, so poll_once still treats 204 as an error.

Closes #4894

Authored with Opus 5.5

[divergentdave]

FWIW status handling in the client and aggregator look good.