Add encode time to query stats

[chencs]

What this PR does

Adds a field encode_time_seconds to query stats log messages, which records the time it takes to encode a response from the frontend.

Which issue(s) this PR fixes or relates to

Fixes #

Checklist

• Tests updated.
• Documentation added.
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX].
• about-versioning.md updated with experimental features.

[dimitarvdimitrov]

Does this include the time it takes queriers to marshal the results? IIRC they use protobuf to talk to the frontend, but I may be wrong. For the purposes of measuring JSON overhead in the query-frontend I think it's better to only measure at the query-frontend.

Would be nice to also get a changelog entry so we document at least somewhere what this new field means.

[chencs]

Does this include the time it takes queriers to marshal the results? IIRC they use protobuf to talk to the frontend, but I may be wrong.

Yeah, the queriers use protobuf to talk to the frontend, and this doesn't include serialization time for that. I think that serialization time (among other things) would be included in the total query_wall_time_seconds, and I think encoding time is not included in query_wall_time_seconds.

For the purposes of measuring JSON overhead in the query-frontend I think it's better to only measure at the query-frontend.

Are you suggesting that we record the time somewhere other than the current changes propose? Sorry if I'm misunderstanding this; I'm pretty sure the current changes are only recording at the query-frontend.

[dimitarvdimitrov]

For the purposes of measuring JSON overhead in the query-frontend I think it's better to only measure at the query-frontend.

Are you suggesting that we record the time somewhere other than the current changes propose? Sorry if I'm misunderstanding this; I'm pretty sure the current changes are only recording at the query-frontend.

oh no, I was just making sure that the changes in this PR don't include the time it takes to marshal the responses in queriers. I think that's the approach we want.

LGTM