feat: json log format and W3C trace context propagation

[joaopaulosr95]

Adds per-request correlation to hsds logs and an optional JSON output format as a way to make it easier to debug it in production.

Motivation

On a busy node, log lines from all in-flight requests interleave with nothing tying a line to the request that produced it, and there's no way to match SN and DN logs for the same client call. OTel-instrumented clients (h5pyd via requests instrumentation included) already send a W3C traceparent header; hsds just needs to read and forward it.

What changed

• log.request() adopts the incoming traceparent (or generates a fresh trace id) into a ContextVar, so every line logged while handling that request carries its trace id
• httpUtil.http_get/post/put/delete forward the header to DNs — one grep finds a client request across all nodes
• new log_format config key: text (default, same REQ>/RSP>/LEVEL> layout as today) or json (one object per line, with method/path/domain on REQ and status/reason/path on RSP lines). Settable via config.yml, LOG_FORMAT env var, or --log_format
• timestamps are now ISO 8601 UTC and on by default (log_timestamps: true)
• removed the dead log-level filter in hsds_app._enqueue_output (was overridden by put_line = True since introduction)

Sample output, same request seen on SN and DN:

sn 2026-07-17T21:28:04.274Z REQ> [aaaabbbbccccdddd0000111122223333] PUT: / [hsdstest/home]
dn1 2026-07-17T21:28:04.292Z REQ> [aaaabbbbccccdddd0000111122223333] POST: /groups

{
  "time": "2026-07-17T21:29:04.832Z", 
  "level": "INFO", 
  "node": "sn", 
  "trace_id": "99998888777766665555444433332222", 
  "method": "GET", 
  "path": "/", 
  "domain": "hsdstest/home", 
  "msg": "GET: / [hsdstest/home]"
}  

Logger API is unchanged (log.info(...) etc.), so no call sites were touched; log_stats/req_count in /info behave as before.

Malformed traceparent headers are ignored, and a fresh trace id is generated.

Testing

New unit tests in tests/unit/logger_test.py cover both formats, level filtering, traceparent parse/generate/reject, and header forwarding on all four http_* helpers against a real aiohttp test server. Also verified on a local SN + 2 DN posix deployment: a single client PUT logs the same trace id on sn/dn1/dn2 in both formats.

[joaopaulosr95]

@jreadey @mattjala flagging you both as you're the top contributors of this project. Yet independent of 447, this is also part of my attempt to make HSDS easier to operate.