Dataplane: Enable dataplane for results of LongToWide

[kylebrandt]

What this PR does / why we need it:

Setting the TypeVersion to greater than [0, 0] (along with Meta.Type being set) indicates that the produced
frame follows the dataplane contract (see https://grafana.com/developers/dataplane/ and https://github.com/grafana/dataplane).

	wideFrame.Meta.TypeVersion = FrameTypeVersion{0, 1}

Okay, so what will that do?
From a user perspective, there are no expected changes.

Certain datasources call LongWide (In particular, SQL datasources) when the query type in the UI is set to "time series". This sets the output of that call to be "dataplane compliant".

Two consumers will follow a different code path in this case (Server Side Expressions and Recorded Queries). From a user perspective things should change unless there is a bug in those code paths.

In SSE, the fork in the code Path is at:

https://github.com/grafana/grafana/blob/2136fd9a929334f39a371c2127c57b6d83abdf28/pkg/expr/converter.go#L32-L38

Special notes for your reviewer:
Since it is possible there could be an issue with a bug in the code that reads data plane data or something overlooked, I'm thinking through some role out possiblities (they have trade offs):

• Just yolo it into the SDK
• Change the signature of the function add argument (maybe Options{} so only done once, but breaking change, so this can be opted into)
• Inverted approach - follow up with opt-out instructions (which would really just be overriding FrameTypeVersion{0, 1} back to FrameTypeVersion{0, 0} after calling LongToWide.

[gabor]

my recommendation would be to go with approach [1] (yolo 😁 ). i mean, it probably won't break and... even if we go with one of the more explicit ways, for example, let's say we go with [2]. this is used in grafana/sqlds, which is used in grafana/google-bigquery-datasource . so, grafana/sqlds will have to update to this change one day, but they cannot really test and make sure it does not break anything anyway, because it'll be grafana/google-bigquery-datasource where the breakage can happen.

that's why i think the simplest approach may be the best here. also, if a datasource sees that things started to break, they can just add a line that deletes the attribute from the dataframe, right?

WDYT?

[kylebrandt]

they can just add a line that deletes the attribute from the dataframe, right?

Yes. Also leaning towards 1, but wanted to give some time for others (like you :-) ) to weigh in.

[gabor]

LGTM 👍