Log trace data for discarded traces

[MarcinGinszt]

What this PR does:
This PR adds the RootSpanName and RootServiceName to log about discarded span.
I added them to main log notifying about this event, since those are basic data allowing identification of the span, and this log doesn't occur often (or, atleast, it should not do so 🙂)

I decided to not include test update at least in the first version of PR, as it looks like they would demand significant changes.
In order to simulate the rootSpanName and rootSpanService, we would need to change the MakeBatch function preparing the spans for tests- first span ParentSpanId would have to be empty. However, definition of "empty" is different between different encodings (empty byte array or simply nil).
As far as I understood, because of this issue testing this feature would demand replicating huge chunks of testing library for different encodings, or additional logic changes. Otherwise, multiple tests from different packages would expect nil and would get []byte{} as a value of first span ParentSpanID- or vice versa.

Which issue(s) this PR fixes:
Fixes #2776

Checklist

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

[CLAassistant]

All committers have signed the CLA.

[joe-elliott]

I decided to not include test update at least in the first version of PR,

Can you extend this test to make sure it extracts the fields correctly?
https://github.com/grafana/tempo/blob/main/tempodb/encoding/vparquet3/compactor_test.go#L150

I think that would be worthwhile. I completely agree it's not worth writing a test to make sure a log line gets written as expected.

Thanks for taking a look at this!

[MarcinGinszt]

@joe-elliott

Can you extend this test to make sure it extracts the fields correctly?
https://github.com/grafana/tempo/blob/main/tempodb/encoding/vparquet3/compactor_test.go#L150

I think that would be worthwhile. I completely agree it's not worth writing a test to make sure a log line gets written as expected.

I'm sorry, I'm afraid I don't know how to do that without replicating huge parts of pkg/util/test/req.go, so that batch is created differently for v1 and v3. I wrote detailed description of the problem in the PR description.
Should I proceed with replicating MakeBatch for different versions?

[joe-elliott]

Should I proceed with replicating MakeBatch for different versions?

Nah, we don't have to do anything complex here.

I'm sorry, I'm afraid I don't know how to do that without replicating huge parts of pkg/util/test/req.go

Is something like this not sufficient:

[joe-elliott]

Nice addition. Thanks!