Query-tee improvements

[charleskorn]

What this PR does

This PR makes a number of improvements to query-tee. The most notable changes are:

• Query-tee now emits trace spans when proxying requests: one span is emitted for each incoming request and another for each outgoing request.
• All log messages associated with a request are tagged with that request's trace ID, just like Mimir itself.
• Query-tee now correctly configures the Jaeger tracing backend, using the same environment variables as Mimir itself.
• Log fields such as path, query and route_name are now set on all log messages emitted while processing a request.
• The user agent is now logged while processing a request.
• All errors encountered while making a request to a backend are logged, such as DNS failures. (Previously, query-tee relied on the comparator to do this, but the comparator is only used when comparing requests.)
• Incoming instant query requests without an explicit value for the time parameter now have this explicitly set for outgoing requests to backends, to prevent comparison failures due to different backends using a slightly different timestamp. This is optional, but enabled by default.
• The microservices Docker Compose environment can now optionally stand up a query-tee instance for testing.

I've tried to make each commit self-contained, and I recommend reviewing each commit separately.

Which issue(s) this PR fixes or relates to

(none)

Checklist

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

[jhesketh]

Nice stuff, thanks! Lgtm, just nits.

[jhesketh]

(nit) could also test where both the form and url params are set

[charleskorn]

Do you mean time is present in both? Or some parameters are in the URL and some are in the body?

[jhesketh]

I meant some parameters are the URL and some in the body. I guess it could also be worth ensuring that if it is set in both that they are passed through unchanged.

[charleskorn]

Added in 527c416.

If time is set in both, then I think all bets are off, so I've left that out for now.

[jhesketh]

If it's set in both I would expect the set values to be passed on. What happens from there I'm unsure, but the former is still something we could check here. Not necessary though imo.

[jhesketh]

Thanks. LGTM